Design of multiple-input power converter for hybrid ... - IEEE Xplore

5 downloads 856 Views 2MB Size Report
School of Electrical and Computer Eng. State University of Campinas ... power required to operate the vehicle is lower than the FC-BU rated power, the UCs can ...
Design of Multiple-Input Power Converter for Hybrid Vehicles L. Solero, A. Lidozzi

J. A. Pomilio

Dept. of Mechanical and Industrial Eng. University ROMA TRE Roma, Italy [email protected]

School of Electrical and Computer Eng. State University of Campinas Campinas, Brasil [email protected]

Abstract— This paper deals with designing and sizing of a Multiple-Input Power Electronic Converter (MIPEC) to be used in an electric vehicle propulsion system that includes a fuel cell (FC) generator and a combined storage unit (CSU). The CSU is composed by an ultracapacitors tank (UC) and a battery unit (BU). MIPEC is responsible for power-flow management onboard the vehicle for each mode of operation. Specifications for MIPEC designing come out from many considerations concerning traction drive and reference driving cycle, on-board power source and storage unit characteristics. However, to date sizing and configuration of both storage units and on-board generators are directly related to traction drive and driving profile (i.e. vehicle performances and characteristics) and no relation with power electronic interface is considered during preliminary design. Then, power electronic interface is selected in order to fit traction drive requirements with power source and storage unit characteristics; as a consequence converter mode of operation lacks of optimization, as well dynamic behavior and efficiency cannot be maximized. In this paper, MIPEC design and power source and storage unit selection are achieved at the same project stage according to traction drive requirements. Experimental results on 60kW power electronic interface are presented. Keywords-ultracapacitors; fuel cell; dc-dc converter; control design

I.

INTRODUCTION

Present researches concerning electric vehicles (EV) and hybrid-electric vehicles (HEV) concentrate in the search for a compact, lightweight, and efficient energy storage system that is both affordable and has acceptable cycle life. The traction system, composed by electric motor, inverter, and associated control circuitry is not the limiting factor to obtain high performance and to permit large-scale production of such vehicles. Attention is now increasingly focused on fuel cell (FC) and hybrid technologies as a way of producing breakthrough vehicles with alternative power plants. A number of auto makers see fuel cell powered vehicles (FC-EVs) as the ultimate route to achieving sustainable long-term alternative propulsion systems. A number of drive-train architectures have recently been proposed to combine two or more on-board generation units (GUs) and storage units (SUs) and to overcome constraints related to fuel consumption, pollution, vehicles’ long distance capability. Interfacing of traction drive requirements with on-board GUs and SUs characteristics and modes of operation calls for suitable power electronic

0-7803-8269-2/04/$17.00 (C) 2004 IEEE.

1145

converter configuration and control. In this paper a 3-inputs/1output converter is proposed for a propulsion system where the GU is a 18kW FC and the combined storage unit (CSU) is formed by lead-acid batteries and ultracapacitors (UCs); however, same converter configuration is appropriate also for different either GUs or SUs. In terms of power sources, the proton exchange membrane FCs are being increasingly accepted as the most appropriate supply for EVs [1,2] because they offer clean and efficient energy without penalizing performance or driving range. A battery storage unit (BU) can be combined with the FC stack to achieve the maximum efficiency for the FC system. The BU delivers the difference between the energy required by the traction drive and the energy supplied by the FC system. In such a system the BU has to deal with power peaks being on demand during either acceleration or braking phases. Such transients result in a hard constraint for BU, which increases the losses and temperature, and reduces its lifetime. Thereby, it is desirable to minimize these power peaks by introducing an additional auxiliary power device: the ultracapacitors [3], which present high power density, obtain regeneration energy at high efficiency during decelerations and supply the stored energy during accelerations. In spite of reaching thousands of Farads, the UCs support very low voltages (1~2.5V). A stack of series-connected UCs can produce an equivalent capacitor of tens of Farads that is able to hold up tens of Volts. The UC stack must supply the power required in excess of the FC–BU system rated power, provided that the UC state of charge (SOC) is greater than a minimum threshold. Whenever the power required to operate the vehicle is lower than the FC-BU rated power, the UCs can be charged with the power in excess. Whenever regenerative braking operations occur, energy is put into the UC tank provided this device is not fully charged yet. The investigated propulsion system arrangement is shown in Fig. 1, where the FC is the main GU and BU and UCs form the CSU. As the FC has poor efficiency at light load, the BU supplies the power at such situation, in order to save total efficiency. The UC tank is used to satisfy acceleration and regenerative braking requirements accomplishing system load transients and improving on-board BU cyclic life. Additionally, it is responsible to control the DC link voltage, while the other sources are current controlled in order to limit the current variation ratio and to prevent excessive peaks.

IUC LINK

SSD UC

VUC

LUC

IUC

UC

ILUC

DSU UC ICUC OUT

CUC OUT

IUC LINK

ICUC IN

SSU UC

DSD UC

CUC IN ILINK IFC LINK

SSD FC

Figure 1. Proposed hybrid drive-train

VFC

FC

The converters can be voltage or current controlled, depending on the source role in the overall system, and their limitations. For example, it is important to limit the current variation in the FC, as well as in the BU. As any capacitor, the UC can be controlled in voltage mode, using a maximum current protection. The reference signals for the control loops are derived from many parameters: the instantaneous load current, the DC link voltage, the BU and UC state of charges, the FC output power, etc. In the following the expressions for reference signals, to be used in MIPEC control, are provided:

1146

ICFC OUT

VLINK

CFC OUT

SSU FC

DSD FC

SSD B

DSU B

CFC IN

IB LINK

MIPEC TOPOLOGY

As mentioned, a Multiple-Input Power Electronic Converter is proposed to interface traction drive requirements with on-board GUs and SUs characteristics. Both FC and UC typically present a lower terminal voltage than the DC voltage necessary to feed the traction inverter. Also for the BU would be of practical interest to use a lower voltage, in order to minimize the series resistance. In such cases it is necessary to use step-up converters for connecting the sources with the common DC bus. Additionally, for the BU and for the UCs it is necessary to have step-down operation in order to recharge them and to accomplish regenerative braking, what means that these converters must be bidirectional in current. A convenient topology is shown in Fig. 2. Each DC-DC converter can be built using a branch of a three-phase DC-AC converter, what means that there are power modules and drives already available in the market. Considering that the common DC link voltage is the highest, the bottom transistor, together with the top diode, configures a boost converter, while the bottom diode and the top transistor realize the buck converter. For the FC converter the buck action must not occur because this apparatus does not take charge from the DC link. A filter capacitor is connected at each source terminals in order to minimize the circulation of high-frequency components through the supplies. This filtering is as effective due to the presence of the sources series resistance.

ILFC

ICFC IN

The goal of this paper is to develop designing and sizing for the DC-DC converters in order to achieve the best compromise for FC generator and CSU sizing, and system dynamic behavior; thus it will be analyzed the influence of the system components in choosing the best feedback variable for each converter and to designing the satisfactory regulators. II.

LFC

IFC

DSU FC

LB

IB

VB B

ILB

ICB OUT

CB OUT

ICB IN

SSUB

DSD B

CB IN

Figure 2. Proposed MIPEC topology

I *BU =

I Lm

I +I ⋅ (1 − d BU ) I*FC = Lm BUc−d 1 − d FC − I FCm ⋅ (1 − d FC ) + I UCc −d ⋅ (1 − d UC ) 1 − d BU VL = const

(1)

where ILm and IFCm are respectively the DC-link and FC measured current, IUCc-d and IBUc-d are the current values of charging and discharging for UC tank and BU whenever storage units’ SOC is either lower or higher of the ordinary admitted range, dBU and dFC are duty cycles of respectively BU and FC converters and VL is the DC-link voltage. First two expressions give reference currents for FC and BU converters, reference signals’ variation is controlled on the basis of generator and storage unit characteristics. UC converter is regulated to keep DC-link voltage either constant or at the most suitable value for traction drive mode of operation. III.

DYNAMIC MODELING

Dynamic modeling is necessary to evince the relationships between system transient behavior and either on-board power source or CSU or traction drive parameters. Small signal modeling, considering the average value of the state variables over one switching period, is a well-known method to analyze non-linear systems, like a switched-mode power supply [4, 5]. The resulting model is valid in a frequency range sufficiently

below the switching frequency. The state equations or the equivalent transfer function can be used to design the regulators in order to achieve a desired system performance.

Substituting (4) into (2) and (3), and neglecting the product of two perturbations, it is possible to obtain the desired transfer function and the output average value:

There are many methods to achieve dynamic modeling of power converters and in this paper the one described in [6] is used. However, some modifications to the well known state variables averaging method have been included in order to take into account possible dependence of output variables from inputs’ vector. The proposed modifications allow to include parasitics of both converter power inductors and capacitors and to take in consideration inner resistance of both vehicle onboard power sources and storage units. Fig. 3 shows the single DC-DC converter considered, including mentioned parasitics of power components. If the converter works as step-up, the average value of the currents ip, iL, and io are positive. In the step-down mode (necessary to recharge BU and UC), the average values are negative. As the dynamic behavior as boost converter imposes more severe restrictions for the control loop design, this case is analyzed at the beginning and, afterwards, the buck operation is verified.

y (s) = [C + G ⋅ A ]⋅ [s ⋅ I − A]−1 ⋅ [( A1 − A 2 ) ⋅ X + (B1 − B 2 ) ⋅ U i ] + (5) d(s) [(C1 − C 2 ) + G ⋅ (A1 − A 2 )]⋅ X + [(E1 − E 2 ) + G ⋅ (B1 − B 2 )]⋅ U i

As the power switches operate in complementary way, the converter always operates in continuous conduction mode (CCM). Notice that, in steady state, the duty-cycle depends only on the voltages Vp and Vo (neglecting the parasitic resistances). The average current is adjusted during the transients and does not depend on the voltages. A. State Variables Averaging Method The state variables, usually the inductors current and the capacitors voltage, are represented in the vector x. The sources are represented in the vector Ui. For the next analysis the sources are supposed of fixed value. For each topologic situation, the differential equations should be obtained and put in the format x& = A 1 ⋅ x + B 1 ⋅ U i . These equations are valid during one topologic combination, for example, while the transistor is on. During the diode conduction, the equations will be x& = A 2 ⋅ x + B 2 ⋅ U i . As the circuit operates in CCM, there are only these two cases.

The same procedure is used to obtain the equations that describe the output variable: y = C1 ⋅ x + G 1 ⋅ x& + E1 ⋅ U i , for the first topologic state and y = C 2 ⋅ x + G 2 ⋅ x& + E 2 ⋅ U i for the second one. The system behavior can be obtained averaging each matrix by the duty-cycle, δ, in which it is valid: x& = [A1 ⋅ δ + A 2 ⋅ (1 − δ) ]⋅ x + [B1 ⋅ δ + B 2 ⋅ (1 − δ) ]⋅ U i

y = [C1 ⋅ δ + C 2 ⋅ (1 − δ)] ⋅ x + [G 1 ⋅ δ + G 2 ⋅ (1 − δ)] ⋅ x& +

[E1 ⋅ δ + E 2 ⋅ (1 − δ)]⋅ U i

(2) (3)

Y = C ⋅ X + E ⋅ Ui

(6)

where A = A 1 ⋅ D + A 2 ⋅ (1 − D) , B = B 1 ⋅ D + B 2 ⋅ (1 − D) , C = C 1 ⋅ D + C 2 ⋅ (1 − D) , G = G 1 ⋅ D + G 2 ⋅ (1 − D) , and E = E 1 ⋅ D + E 2 ⋅ (1 − D) . B. Boost-Converter Let us consider the boost converter, in the CCM, having a capacitive input filter and including parasitics of power inductors and capacitors. The voltage source presents a series resistance R. The load is represented by a current source that, for the dynamic analysis, is an additional input. Fig. 3 shows the circuit and Fig. 4 indicates both equivalent topologies.

Taking the voltage vout as the output variable, the transfer function to the duty-cycle, that is the control variable, is calculated using the following equations: v o    x =  vi   i L 

,

 Io  Ui =   Vp 

,

C1 = [1 0 0]

C2 = [1 0 R o ]

,

E1 = E 2 = [− R o

0]

G1 = G 2 = [0 0 0]

 1   1  0 0 −  −  C C o o     1 1     B1 =  0  , B 2 =  0 C (R + R )  + ( ) C R R i i  i i    Ri Ri  Ro   0   L L(R + R )   L(R + R i )  i   

,

,

  0 0 0    1 R   − A 1 = 0 − , C i (R + R i ) C i (R + R i )   R ⋅ Rl + Rl ⋅ Ri + Ri ⋅ R  R 0 −   L(R + R i ) L(R + R i )   0   A2 =  0  − 1  L 

0 −

1 C i (R + R i ) R L(R + R i )

 1  Co  R  −  C i (R + R i )  R ⋅ Rl + Rl ⋅ Ri + Ri ⋅ R + R ⋅ Ro + Ro ⋅ Ri  −  L(R + R i ) 

.

Thus, the resulting transfer function is: v out (s) = C ⋅ [s ⋅ I − A ]−1 ⋅ [( A1 − A 2 ) ⋅ X + (B1 − B 2 ) ⋅ U i ] + (C1 − C 2 ) ⋅ X d(s)

(7)

where, in case of ideal converter, B1=B2 and C1=C2 [7].

It is possible to split the state variables, the output and the control variable (duty-cycle) in their average value plus a perturbation: x = X+x y=Y+y (4) δ=D+d

1147

This expression is used to analyze UC converter dynamic behavior; as it will be more clearly detailed in the next paragraph, the behavior of the resulting transfer function is different from ideal converter transfer function mainly for the presence of an additional zero; which is caused by output capacitor resistance and capacitance (R0, C0) and it is usually positioned at quite high frequency.

FC and BU converters’ dynamic behavior can be investigated by using either (8) or (9). However, control of the source current allows at least a reduced order filter in the feedback path; thus, the resulting higher phase margin of the converter transfer function tolerates high gain regulators and dynamic of the converter is improved. IV.

Power sizing of both generation and storage units has been achieved on the basis of standard driving cycle and desired performances for a small size HEVs class. The chosen generation unit is a 18kW FC since it is able to deliver the average power required by the most common combined urbanhighway driving cycles. Efficiency of FC generator dramatically decreases when required power is less than 1015% of the FC maximum deliverable power; therefore, the FC generator should be switched off and the BU is in charge of supplying the required power. Total energy required for the BU, considering that it is also responsible of on-board electric loads, is almost 7000kJ; besides the BU itself should be able to deliver 10kW for at least 10 minutes in order to assure 130km/h cruising speed of the vehicle for the mentioned time. A 30kW – 260kJ UC tank is needed as it is responsible of great part of vehicle accelerations and regenerative brakings, in fact current variations are limited for both FC generator and BU to reduce stress and to assure them a sufficient life-time.

Figure 3. DC-DC converter for dynamic modeling

a

b Figure 4. (a) Low-side switch conduction (SSU or DSD); (b) high-side switch conduction (DSU or SSD).

As FC and BU converters are controlled in current mode, it is necessary to define in which point the current should be controlled. The inductor current and the source current are the two options which have been investigated. When the inductor current is the controlled variable an additional low-pass filter in the feedback path is required for low inductance value, thus making quite difficult to design a regulator for having wide compensation band with secure phase margin. In case of source current as controlled variable, the natural filtering introduced by the input capacitor avoids the additional low-pass filter and regulator design has less constraints. Taking the current iL as the output variable, no differences are found in A and B matrices since they depend on circuit modeling, whereas C, G, and E are the following: C1 = C 2 = [0 0 1] , G 1 = G 2 = [0 0 0] , E1 = E 2 = [0 0] ; thus, the resulting function to the duty cycle is different from ideal converter situation only for B1≠B2 and it can be expressed as iL (s) (8) = C ⋅ [s ⋅ I − A ]−1 ⋅ [( A1 − A 2 ) ⋅ X + (B1 − B 2 ) ⋅ U i ] d(s)

When the current ip is the output variable matrices C, G, and E are the following: C1 = C2 = [0 0 1] , G1 = G 2 = [0 Ci 0] , E1 = E 2 = [0 0] ; the resulting function to the duty cycle is different from the ideal converter case for the presence of the G matrix and because B1≠B2: ip (s)

= [C + G ⋅ A] ⋅ [s ⋅ I − A ]−1 ⋅ [( A1 − A 2 ) ⋅ X + (B1 − B 2 ) ⋅ U i ] + d(s) G ⋅ (A1 − A 2 ) ⋅ X + G ⋅ (B1 − B 2 ) ⋅ U i

MIPEC DESIGN

(9)

1148

The commercial traction drive used in the proposed propulsion system is formed of a VSI-inverter and an induction motor, and it is rated 216V – 140A at DC-link. Selection of voltage and current rated values for both generation and storage units must comply with traction drive and MIPEC topology specifications, in fact the number of elements that must be connected in series to form either the FC generator or the BU or the UC tank has minimum and maximum values related to both DC-link voltage and acceptable duty cycle values for MIPEC switches. The following expressions can be used to find out the most suitable number of elements to be connected in series: n ⋅ (Vel − R e ⋅ I M ) ≥ Vlink ⋅ (1 − d SUh )

n ⋅ (Veh − R e ⋅ I m ) ≤ Vlink ⋅ (1 − d SUl )

(10)

n ⋅ (Vel + R e ⋅ I m ) ≥ Vlink ⋅ d SDl

(11)

n ⋅ (Veh + R e ⋅ I M ) ≤ Vlink ⋅ d SDh

where Vel and Veh are the lowest and highest acceptable element voltage, Re is the equivalent inner resistance for each element, Im and IM are the minimum and maximum accepted values for each unit under investigation, dSUl and dSUh are respectively the lowest and highest value for switch duty cycle in boost mode of operation, whereas the same meaning is related to dSDl and dSDh in buck mode of operation. Investigation on commercial products and the iterative applying of (10) and (11) for both UC tank and BU, and only (10) for FC generator, led to define generation and storage units’ configuration as shown in Table I. Final FC generator configuration is rated 18kW - 160A, 112V (inner voltage drop is included) at rated power – and elements number of 200 was chosen among the results satisfying (10) (i.e. 135≤n≤203) in order to limit dSUh at 0.5 in

steady state condition and improve the switch utilization factor. Similar considerations on improving switch utilization factor led to choose 12 elements for BU − 9≤n≤13 is the solution to (10) and (11) – that is formed of Genesis batteries rated 12V, 13Ah. Cost saving and energy specification for UC tank affected the choice of UC modules number; in fact the best solution for (10) and (11) is n=4 on the basis of the switch utilization factor. However, on the basis of commercial products available in the market it would result in over-sizing the UC unit energy, thus 3 modules of Maxwell UCs each rated 42V, 145F have been considered. TABLE I. Specs Vlink [V] dSUh dSUl dSDh dSDl

current increases, as shown in Fig. 6. Bode diagrams of transfer functions for inductor current iL and input power unit current ip are respectively shown in Figs. 7 and 8. Modeling of non ideal components doesn’t affect at all the resulting transfer function when iL is the output variable, whereas an additional zero is present for the ip case. Also in this case the additional zero is at high frequency and it is related to the product ESR·Ci.

GENERATION AND STORAGE UNITS CONFIGURATION Values 216 0.65 0.1 0.9 0.1

Paramet. Re [mΩ] VeM [V] Vem [V] IM [A] Im [A] n

FC elem. 2.75 1.0 1.0 160 16 200

BU elem. UC module 13 15 14 45 10 28.9 70 250 0 0 12 3

Traction drive DC-link voltage and current values of generation and storage units represent specifications for switching components’ selection. Voltage ripple in DC-link is the parameter used for MIPEC output capacitance sizing, however it must be ensured that selected capacitor tolerates RMS value of the output ripple current [8]. Input capacitor and inductor are responsible of input current ripple reduction: choice of 15kHz as switching frequency led to components’ selection as shown in Table II. TABLE II.

MIPEC POWER COMPONENTS PM300DSA060 (IPM 2-pack mod.) 300 600

Power Semiconductors Rated Current [A] Rated Voltage [V] Inductors Inductance [µH] Rated Current [A] ESRL [mΩ] Capacitors Capacitance [mF] Rated Voltage [V] ESRC [mΩ]

Figure 5. UC converter Bode diagram: output voltage Vout to duty cycle

LFC 130 160 9.6 COUT 15 385 5

LUC 52 200 2.4 CFC-IN 1 385 74

Figure 6. UC converter: RHP zero frequency vs. load current LBU 160 80 9.8

CUC-IN 2 385 37

CBU-IN 1 385 74

Parameters’ values shown in Tables I and II are used in transfer functions (7), (8) and (9) for MIPEC dynamic investigation. It is found that output capacitors’ equivalent series resistance (ESR) introduces a high frequency additional zero in the output voltage transfer function with respect to ideal converter case as it can be noticed in Fig. 5; being the frequency position of the additional zero inversely related to the product ESR·Cout, significant values of the output capacitance (decrease of ESR is not linear with capacitance increase) affect regulators design since the additional zero frequency is lowered. Influence of the additional zero is significant in design of regulators for UC converter, which is supposed to react with very high dynamic to DC-link variations. Poles and zeroes values of the output voltage transfer function depend on converter operation point (average duty-cycle and average load current). One of the zeros is at the right half-plane (RHP) and its value decreases as the output

1149

Figure 7. UC converter Bode diagram: inductor iL current to duty cycle

Figure 8. UC converter Bode diagram: power source current ip to duty cycle

Extensive investigation on sensitivity of non ideal components in the MIPEC model resulted that converter input capacitance and inductance form a resonant path which affects regulators design when input power units have not negligible equivalent inner resistance (i.e. large number of series connected elements). In order to reduce the resonance, the number of series connected elements should be restricted; in particular for FC generator and BU, such a requirement is in conflict with appropriate sizing of input power units, then maximum tolerable resonance amplitude should be taken into account at definition of the series elements’ number. V.

SIMULATION AND EXPERIMENTAL RESULTS

Current loop with PI type regulator is chosen for both FC and BU power stage regulation in order to directly control each source current; measured currents are filtered by means of Butterworth 2nd order continuous-time active filter to cut off switching ripple when inductor current is the output variable. UC power stage is devoted to dc link voltage control, thus a configuration with outer voltage loop plus inner current loop is proposed for the investigated application; the current loop has come out to be indispensable to control current either fed or soaked by UC tank whenever any DC-link voltage unbalance occurs. A well-designed fast control loop can greatly improve the whole system dynamic performance during load transients, as well the propulsion system makes the best use of UCs own dynamic characteristics and high power density. To this purpose a PI type regulator is chosen for the current loop, whereas a 2-zeros/3-poles regulator has been designed for the voltage loop [7]. Control loops’ design has been investigated by means of Matlab-Simulink models in which quantization of measures and control discrete transfer functions have been taken in consideration as well as both true calculus mode adopted on DSP and control loop delays have been included. Dynamic response and stability for each converter included in MIPEC configuration have been tested at different reference signals and load variations, achieved results show good dynamic performance in every simulated operating condition. In Figs. 9 and 10, UC converter response is shown for load current step variation of 30A respectively when iL and ip are used as regulated variable in current loop. Both simulations show a smooth regulation of the output voltage; however, the mode of operation of a Butterworth 2nd order continuous-time active filter is considered when inductor current is controlled, whereas no filtering of the controlled current is required to control the power source current. A fixed-point 16-bits DSP from Analog Devices has been used in order to implement the MIPEC whole control system. MIPEC switching frequency is chosen to be 15kHz according to hardware components specifications; besides, a fixed length of 133µs is chosen as maximum period required for the whole control algorithm to be completely executed. As a consequence, the maximum achievable sampling frequency of 7.5kHz has been chosen to implement regulators’ transfer functions in discrete form for both BU and FC current loops; whereas the frequency of 1.875kHz has been used for the UC double loop. In fact, a 4 times reduced frequency improves stiffness of the UC control system by reducing the effects of the voltage loop RHP-zero. DSP standard fixed point

1150

configuration was adopted for the whole algorithm except for UC regulator implemented by using the emulated floating point mode of operation. Control algorithm includes time-variation limiting of currents should be supplied by either FC generator or BU to achieve a safe dynamic mode of operation for both the power source and the main energy storage. UC power stage currents have no time-variation limitation in order to achieve the most effective output voltage regulation. DSP is also responsible for system protection actions (i.e. overcurrents, overvoltages and overtemperature). Experimental measurement revealed the DSP takes 110µs to complete the control algorithm execution, this value well fits the previously chosen maximum available time.

Figure 9. UC converter response (vout and iL) at load current step variation

Figure 10. UC converter response (vout and ip) at load current step variation

For experimental testing activity purpose, FC generator has been simulated by means of a 20kW regulated DC power source, UC tank was accomplished with 3 BMOD0115A09 Maxwell modules (42V, 145F) in series connections, series connection of 12 modules from Genesis (12V, 13Ah) form the BU. Dynamic response for each power stage converter has been tested; transients of UC power stage at step variation of voltage reference have been investigated in detail at no-load operation, which is the worst case for output voltage control as MIPEC output capacitors provide a very low damping effect. Load testing has been carried out by operating all MIPEC power stages at the same time. Figs. 11 and 12 show MIPEC dynamic performance at load operation corresponding to resistive load step variation respectively from no-load to 4.33Ω (i.e. 50A at 216V) and from 4.33Ω to no-load; current variation vs. time limitation is implemented for both FC generator and BU storage, UC converter acts in order to compensate the output voltage variation. In case of quite low SOC for storage units, as transient is completed, BU and UC would require to be charged (at constant current) from FC generator.

Figure 11. MIPEC load transient operation: 0-50A load current step variation

Figure 13. Propulsion system transient operation: traction motor torque step

Figure 12. MIPEC load transient operation: 50-0A load current step variation

Figure 14. Propulsion system testing: modified urban ECE-15 driving cycle

Complete propulsion system has been loaded with several different driving cycles and tested at ENEA lab facilities. Fig. 13 shows the current waveforms for each input power source and DC-link when almost 50Nm torque step is applied to the traction motor; whereas Fig. 14 depicts same current waveforms when the complete propulsion system is loaded with urban ECE-15 driving cycle, whose accelerations have been 40% increased in order to achieve a more realistic testing. It can be noticed that UCs run during short and severe both accelerations and brakings, whereas gentle speed variations are accomplished by FC generator.

accomplished at a suitable HEV test-bed where applied traction drive torque transients and several driving cycles proved MIPEC good dynamic behavior.

VI.

CONCLUSIONS

The topology for a 3-inputs/1-output power converter (MIPEC) devoted to HEVs applications has been presented. On-board generation and storage units’ sizing as well MIPEC design and prototypal realization have been carried on according to specifications from traction drive and vehicle performances. Dynamic modeling of the proposed converter has been achieved to evince dependence of system transient behavior from parameters of on-board generator, storage units and traction drive. Simulations confirmed performances of designed regulators and control strategy. A 60kW MIPEC prototype has been used for the hybrid propulsion system where a 18kW fuel cell is the main generation unit and batteries and ultracapacitors form the combined storage unit. Experimental testing of the whole system has been

1151

REFERENCES [1] [2] [3]

[4]

[5]

[6]

[7]

[8]

V. Raman: “The Hydrogen Fuel Option for Fuel Cell Vehicle Fleets,” Fuel Cell Power for Transportation, SAE SP-1425, 1999. K. Dircks: “Recent Advances in Fuel Cells for Transportation Applications,” Fuel Cell Power for Transport., SAE SP-1425, 1999. A. Rufer, P. Barrade: “Key Developments for Supercapacitive Energy Storage: Power Electronic Converters, Systems and Control,” Second Boostcap Meeting, Ecole d'Ingénieurs de Fribourg, Fribourg, Switzerland. A. Di Napoli, F. Crescimbini, F.G. Capponi, L. Solero: “Control Strategy for Multiple Input DC-DC Power Converters Devoted to Hybrid Vehicle Propulsion System,” Proc. of IEEE ISIE 2002, L’Aquila, Italy, July 2002, pp. 1036-1041. R. D. Middlebrook and S. Cuk: “A General Unified Approach to Modeling Switching Converter Power Stage,” Proc. of the IEEE PESC 1976, pp. 18-34. P. T. Krein, J. Bentsman, R. M. Bass and B. L. Lesieutre: “On the use of averaging for the analysis of power electronic systems,” IEEE Trans. on Power Electronics, vol. 5, no. 2, Apr. 1990, pp. 182-190. J.A. Pomilio, L. Solero, F. Crescimbini, A. Di Napoli, “Dynamic Modelling and Regulators Design for Multiple Input Power Converters for the Propulsion System of Electric Vehicles,” Proc. of the COBEP’2003, Fortaleza – Ceara (Brazil), Sept. 2003, pp. 362-367. A. Di Napoli, F. Crescimbini, S. Rodo, L. Solero: “Multiple Input DCDC Converter for Fuel-Cell Powered Hybrid Vehicles,” Proc. of the IEEE PESC 2002, Cairns, Australia, June 2002.