This action might not be possible to undo. Are you sure you want to continue?
2, JUNE 2007
Control of a Fuel Cell Based Z-Source Converter
Jin-Woo Jung, Member, IEEE, and Ali Keyhani, Fellow, IEEE
Abstract—This paper presents system modeling, modiﬁed space vector PWM implementation (MSVPWM), and control system design of a Z-source converter. For an electrical analysis of the fuel cell powered systems, the fuel cell system is modeled by an R–C circuit including its voltage–current polarization characteristics. A discrete-time state space equation is given to implement digital control and a space vector pulse-width modulation technique is modiﬁed to realize the shoot-through zero vectors that boost the dc-link voltage. Three discrete-time controllers and an asymptotic observer are designed. Index Terms—Digital control, distributed generation systems, dynamic model, fuel cells, space vector PWM, Z-source converter.
Conﬁguration of the fuel cell systems.
ECENT technological advances in small generators, power converter technology, and energy storage devices have provided a new opportunity for distributed generation systems (DGS). The DGS can offer improved service reliability, better economics and a reduced dependence on the local utility. Environmental friendly emerging technologies such as wind turbines, hydro turbines, photovoltaic cells, and fuel cells have also encouraged a more decentralized approach to power delivery due to environmental regulations regarding greenhouse gas emission –. Especially, the fuel cells can offer the prospect of supplying the world with safe, clean, efﬁcient, sustainable electrical energy because they use hydrogen as a fuel and can continuously generate electric power as long as hydrogen and oxygen are available. Furthermore, the fuel cell development has been accelerated by investment of the automotive industry since it has an interest in clean energy for transportation. A fuel cell is an electrochemical device that converts chemical energy directly into electric energy by reaction of hydrogen from the fuel and oxygen from the air. The fuel cells also produce water and heat by combining hydrogen with oxygen, and have 50% efﬁciency for only electricity but could reach 85% in the case of combined heat and power (CHP). Also, the fuel cells can be used for various applications such as stationary sites (buildings, hospitals, domestic utility, etc.), transportation (fuel cell vehicle), portable power (laptop, cell phone), and distributed power. For practical realization of the fuel cell powered systems, slow dynamics of the fuel cells should be considered. Based on complicated physical and chemical processes, mathematical dynamic models of the fuel cells have been presented – for
Manuscript received September 24, 2005; revised September 24, 2005. This work was supported in part by the National Science Foundation under the grant NSF ECE 0501349. Paper no. TEC-00305-2004. J. W. Jung is with the Samsung SDI Co., Yongin, Gyeonggi-Do 449-902, Korea (e-mail: firstname.lastname@example.org). A. Keyhani is with the Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH 43210 USA (e-mail: keyhani.1@ osu.edu). Digital Object Identiﬁer 10.1109/TEC.2006.874232
design of fuel cells. For a simpliﬁed dynamic analysis, a secondorder model  or a ﬁrst-order time delay circuit  has been proposed. For residential and automotive applications, the design and control of dc-to-dc boost converters and dc-to-ac inverters have been investigated to design low-cost and small-sized power converters –. However, the fuel cells are modeled by a dc voltage source and single-phase inverters are considered. In , , a Z-source converter that can boost a dc-link voltage by employing L–C impedance components and a shootthrough without any boost converter is proposed but it shows its operation under only a linear/heavy load and an open-loop control without considering the dynamics of the fuel cell system. In addition, techniques to produce an ac output voltage with low THD in a three-phase PWM inverter have been proposed –. Real-time deadbeat controllers – have low THD for linear load and a fast transient response for load disturbances. However, it is known that these techniques have a high THD under nonlinear loads. In this paper, system modeling, modiﬁed space vector PWM (MSVPWM) implementation, and control system design of a Z-source converter are presented. For considering dynamics of the fuel cell system, the fuel cell is modeled by an R–C circuit including its voltage–current polarization characteristics. A space vector PWM technique is modiﬁed to realize the shoot-through zero vectors that boost the dc-link voltage. Three discrete-time controllers and an asymptotic observer are designed to guarantee a fast, no-overshoot current response, a zero steady-state voltage and low THD under heavy/light loads, linear/nonlinear loads, and load changes. To verify the proposed control method, simulation studies using Matlab/Simulink are performed in a three-phase ac 208 V/60 Hz/10 kVA under various conditions. II. MODELING OF FUEL CELL In the case of indirectly producing the hydrogen by the reformer, the fuel cell generation system consists of three parts: a reformer, stack, and power converters (Fig. 1). The reformer
0885-8969/$25.00 © 2006 IEEE
3. and Rs and Cs . The PWM inverter DSP controller communicates with the fuel cell processor to equalize power available from the stack to power requested by the load. As described in Fig. In this paper. the maximum cell current and stack voltage increase as the hydrogen mass ﬂow rate increases. The stack has many unit cells that are stacked in series or parallel to generate a higher voltage needed for their applications because one single cell that consists of electrolyte. The ﬁlter capacitors (Cf ) to ﬁlter . 4. NO. Voltage–current characteristics of a cell. Fig. 5. Fig. For dynamic modeling of the fuel cells. Fig. a PWM inverter DSP controller. 2 shows a block diagram of reformer and stack to illustrate the generation of dc power. a Z-source converter. controls the Z-source converter. an R–C circuit model is used to realize slow dynamics caused by a chemical/electrical response of the reformer and stack as depicted in Fig. JUNE 2007 Fig. Fig. the shoot-through zero vector) besides eight switching vectors. On the other hand. In Fig. Fig. Also it generates dc electric power by an electrochemical reaction of hydrogen and oxygen. The controller monitors the stack current and voltage to assure the proper operation of the fuel cell. The Z-source converter utilizes the shootthrough to directly step up a dc source voltage without a boost dc/dc power converter. 3. stack. the Z-source converter has one more vector (i. 5. the fuel processor controls the reformer to produce hydrogen for power requested from the PWM inverter DSP controller. 5 shows the total system diagram of a fuel cell based Z-source converter that consists of a reformer. 2.e. The power converters convert a low voltage dc from the fuel cell to a high voltage dc or a sinusoidal ac. The traditional three-phase VSI has eight switching vectors that consist of six active vectors (V1 –V6 ) and two zero vectors (V0 . the output of the stack shows a family of voltage–current curves for various hydrogen mass ﬂow rates. produces hydrogen gas from fuels and then supplies it to the stack. the V –I characteristics shown in Fig. and a three-phase load. Electrical equivalent circuit model of the fuel cell. VOL. Dynamic response of the stack is considered to have faster response due to the electrochemical process of hydrogen and oxygen compared to that of the reformer. III. 6 shows a detailed system conﬁguration of a fuel cell powered Z-source converter. In the conventional three-phase voltage source inverter (VSI). and in this study. and senses output voltages/currents for a closed-loop control. V7 ). the reformer and stack are modeled by Rr and Cr . the reformer and stack that determine the dynamic response of the fuel cell systems are further described. to investigate an overall operation of fuel cell powered systems. Fig. 22. The voltage–current polarization characteristics of the stack is also considered. 2. respectively. the dynamics of the reformer may be represented by a second-order model  or a ﬁrst-order time delay model . the dynamic response of the reformer and stack and a cell current voltage curve need to be modeled for more realistic analysis of the fuel cell systems. 4 is used. As a result. CONFIGURATION OF Z-SOURCE CONVERTER FOR FUEL CELL SYSTEMS The Z-source converter is based on a new concept different from a conventional dc-to-dc or dc-to-ac power converter . a fuel processor controller.468 IEEE TRANSACTIONS ON ENERGY CONVERSION.. 2. a boosted voltage rate depends on total duration (Ta ) of the shoot-through zero vectors over one switching period (Tz ). Block diagram of reformer and stack. and plates. separators. Thus. produces approximately 0. In this ﬁgure. That is. The reformer affects dynamic response of the fuel cell system because it takes up several to tens of seconds to convert the fuel into the hydrogen depending on the demand of the load current. System diagram of a fuel cell based Z-source converter.7 V dc. the shoot-through in which both power switches in a leg are at once turned on must be avoided because it causes a short circuit. Thus.
the output voltage (Vin ) of fuel cell. System Modeling Fig. 7 shows a simpliﬁed circuit model of Fig. IV. the system model (1) in the abc reference frame can be transformed to (3) in the stationary dq reference frame as follows: 1 dVLdq 1 = Iidq − Tidq ILdq dt 3Cf 3Cf dIidq 1 1 = − VLdq + Vidq dt Lf Lf where Tidq = [Ks Ti K−1 ]row. The diode between the fuel cell and Z-source converter is required to prevent a reverse current that can damage the fuel cell. and iiC . the system consists of a fuel cell. respectively. C1 . Detailed conﬁguration of the Z-source converter system. the load line-to-line fdq0 = [fd fq f0 ]T . . The circuit model described in Fig. an output ﬁlter (Lf and Cf ). a diode. 4×4 B= E= 02×2 I2×2 . and impedance components (L1 . the above state equations can be transformed from the abc reference frame into stationary dq reference frame that consists of the horizontal (d) and vertical (q) axes. 1 − 3Cf Tidq 02×2 4×2 Note that the line-to-line load voltage VLdq and inverter output current Iidq are the state variables of the system. Also. a vector is deﬁned as Ii = [iiAB iiBC iiCA ]T = [iiA − iiB iiB − iiC iiC − iiA ]T . out harmonics of the inverter output voltage due to the PWM technique are added to the conventional Z-source converter. In Fig. 7 shows a system consisting of a dc voltage source. Using (2). and a threephase load. Notice that a fuel cell. an output ﬁlter (Lf and Cf ).1. which is also equal to the voltages (VC1 and VC2 ) of two capacitors (C1 and C2 ). and line-to-line load voltage (VL ) are measured to implement the feedback control.2 = s 3 2 1 1 √ 3 −1 √ 3 (3) 1 . and Vdc denotes the average dc-link voltage. Fig. (1) 0 −1 . L2 . inverter output current (Ii ). SYSTEM MODELING/PWM IMPLEMENTATION A. the inverter . dt Lf Lf where 1 −1 Ti = 0 1 −1 0 Fig.JUNG AND KEYHANI: CONTROL OF A FUEL CELL BASED Z-SOURCE CONVERTER 469 voltage and phase current vectors can be represented by VL = [VLAB VLBC VLCA ]T and IL = [iLA iLB iLC ]T . a three-phase inverter. C1 . and f denotes either a voltage or a current variable. 7 uses the following quantities. and a three-phase load (RL ). d = [ILdq ]2×1 . 6. impedance components (L1 . 4×2 u = [Vidq ]2×1 . and the threephase inverter output currents are iiA . 7. a diode (D). The relation between these two reference frames is below fdq0 = Ks fabc where 1 2 Ks = 0 3 1/2 −1/2 √ 3/2 1/2 −1/2 √ − 3/2 . 4×1 A= 02×2 1 − Lf I2×2 I2×2 02×2 . iiB . Based on these currents. 6. capacitor voltage (VC2 ). The inverter output line-to-line voltage is represented by the vector Vi = [ViAB ViBC ViCA ]T . and C2 ). 1/2 (2) Fig. and C2 ) are replaced with a dc-link voltage source (Vdc ) for circuit modeling. 1 To implement the space vector PWM. Simpliﬁed circuit model of the Z-source converter. a three-phase inverter. Also. L2 . The given plant model (3) can be expressed as the following continuous-time state space equation ˙ X(t) = AX(t) + Bu(t) + Ed(t) where X= VLdq Iidq 1 Lf 1 3Cf (4) . The L–C output ﬁlter yields the following state equations by KCL and KVL: dVL 1 1 = Ii − Ti IL dt 3Cf 3Cf 1 dIi 1 = − VL + Vi . 6 for an analytic modeling of the Z-source converter using fuel cells. fabc = [fa fb fc ]T .column.
(c) Sector 3. NO. For implementation of the modiﬁed space vector PWM. and T0 ) of the traditional SVPWM in order to boost the dc-link voltage of the Z-source converter and to generate the sinusoid ac output voltage. respectively. 8 shows both the conventional and modiﬁed switching patterns for the Z-source converter at each sector. (b) Sector 2. TABLE I SWITCHING TIME DURATION AT EACH SECTOR output line-to-line voltage Vidq is the control input (u). the peak value of inverter output voltage (Va p ) deﬁnitely depends on both the modulation index (M ) and the boost factor (K). JUNE 2007 Fig. VOL. the conventional SVPWM technique should be modiﬁed.470 IEEE TRANSACTIONS ON ENERGY CONVERSION. the peak value of phase-a inverter output voltage can be written as  Va where Tz 1 K= = ≥1 Tb − Ta 1 − 2 Ta Tz K is called a boost factor and M and Vp dc denote the modulation index and the peak dc-link voltage. (a) Sector 1. T2 . 2. (f) Sector 6. In this ﬁgure. Modiﬁed SVPWM implementation. (d) Sector 4. and the load current ILdq is deﬁned as the disturbance (d). 22. From (5). p =M Vin Vp dc = MK 2 2 (5) Tz = Ta + Tb . Space Vector PWM Implementation For realization of the Z-source converter to utilize the shootthrough. (e) Sector 5. and Tb is the total duration of nonshoot-through switching vectors during Tz . and the boost factor (K) is determined by . B. Ta is the total duration of shoot-through zero vectors during Tz . 8. the switching time of the upper switches and lower switches in a three-phase inverter is summarized in Table I below. In the Z-source converter. a new duration (T ) should be added to the switching time (T1 . Tz is the switching period. Fig.
Using the asymptotic observer which estimates the load current (ILdq ) deﬁned as the disturbance. Discrete-Time Sliding Mode Current Controller A discrete-time sliding mode control (DSMC) with the asymptotic observer is used for current regulation in an inner loop because of the fast and no-overshoot response . (9) Xb = [Iidq ]. a sliding mode manifold can be selected below s(k) = y1 (k) − y1 ref (k) = C1 X(k) − y1 ref (k). The control law that satisﬁes (9) and yields motion in the manifold s(k) = 0 is called “equivalent control” and is given by ˆ ueq (k) = (C1 B∗ )−1 (I∗ (k) − C1 A∗ X(k) − C1 E∗ d(k)). With the information of the disturbance obtained from the observer. V. discrete-time sliding mode can be reached if the control input u(k) is designed as the solution of s(k + 1) = y1 (k + 1) − y1 ∗ ref (k ∗ + 1) ˆ = C1 A X(k) + C1 B u(k) + C1 E∗ d(k) (6) − y1 ref (k + 1) = 0. Even if the output voltage of inverter and dc-link voltage can be controlled by adjusting Ta . ˆ d = [ˆLdq ]. the maximum available shootthrough interval (Ta ) to boost the dc-link voltage (vi ) is limited by the zero vector duration (T0 /2) which is determined by the modulation index (M = a(4/3)). Total control system block diagram. idq C1 = B∗ = 0 0 0 1 0 . A∗ a =e Aa Tz A∗∗ = 0 Tz eAa (Tz −r ) Aa dτ. A. Thus. the boosted rate of the dc-link voltage is dependent on the total duration (Ta = 3T ) of shoot-through zero vectors that simultaneously turn on both power switches in a leg. the load current can be estimated as follows: ˆ ˆ Xa (k + 1) = A∗ Xa (k) + A∗ Xb (k) − A∗∗ Xa (k) a a b ˆ ˆ d(k) = ˆLdq (k) = L1 (Xa (k) − Xa (k)) I where ˆ ˆ Xa = [VLdq ]. CONTROL SYSTEM DESIGN Fig. and has only one shoot-through zero state (T ) over one period (Tz ) in any sector without the change of total zero vectors (V0 . eA(Tz −τ ) Edτ. 3Cf L1 = observer gain.JUNG AND KEYHANI: CONTROL OF A FUEL CELL BASED Z-SOURCE CONVERTER 471 Fig. V7 . In the following sections. 0 0 0 1 Tz A∗ = eATz . (8) Therefore. Aa = − L1 Tidq . In order to control the output y1 (k) to follow the reference y1 ref (k). and T ) and total nonzero switching vectors (V1 – V6 ). the discrete-time state space equation of the system (4) for design of a current controller can be expressed as ˆ X(k + 1) = A∗ X(k) + B∗ u(k) + E∗ d(k) y (k) = C1 X(k) 1 eidq (k) = y1 (k) − y1 ref (k) where y1 = [Iidq ]. 3Cf Ab = 1 I2×2 . Moreover. a ratio Ta /Tz . In Fig. sum of the modulation index (M = Tb /Tz ) and the ratio Ta /Tz is always equal to unity. Ab dτ. idq (10) . 8. 9. = 0 e Aa (Tz −r ) . 9 shows a block diagram of the control system proposed for the fuel cell based Z-source converter. eA(Tz −τ ) B dτ. each phase still switches on and off once per switching cycle (Tz ). A∗ b Tz Xa = [VLdq ]. As shown in this ﬁgure. I E∗ = 0 Tz (7) y1 ref = [I∗ ]. the overall system has three feedback controllers and an asymptotic observer. design of each controller will be described in detail.
c B∗ = c 0 Tz With control law (11). ∗ ∗ ∗ ∗ (14) eAc (Tz −τ ) Bc dτ. ±jω2 . the discrete-time sliding mode can be reached after a ﬁnite number of steps and the control voltage limit u0 is also determined by the SVPWM inverter. 2. 0 ˆ A= Ad −B∗ Cd c 0 . . A∗ c 0 . ±jω3 . I The stabilizing compensator which yields the control signal u1 ensures the stability and desirable performance of the overall system through a feedback gain K which minimizes a discrete linear quadratic performance index ∞ Cd = [I2×2 ∗ −1 ∗ −1 02×2 ]. Ac = 04×4 Ac2 04×4 04×4 02×2 Ac3 12×12 Bc1 02×2 I2×2 Bc = Bc2 .idq ) is limited by maximum current to protect the system under over-current. Since the dynamics of the DSMC is included in the inner loop. an augmented system combining both the plant (12) and the servo-compensator (14) is ˆ ˆ ˆ ˆ ˆˆ ˆ X(k + 1) = AX(k) + Bu1 (k) + E1 d(k) + E2 yd ref (k) (15) where X(k) ˆ X(k) = . 22. representing sinusoidal signals with fundamental frequency ω1 and harmonic frequencies ω2 . a discrete-time voltage controller that consists of a servo-compensator with the internal model principle and a stabilizing compensator with the optimal control is used for voltage regulation in an outer loop due to its capability to perform zero steady-state tracking error under unknown load and to eliminate harmonics of any speciﬁed frequencies with guaranteed system stability . its model has to be combined with the original plant. (12) yd ref (k) ∗ = VLdq (k). we assume the tracking/disturbance poles are ±jω1 . and ±jω3 . Ed = E − B (C1 B ) C1 E .idq (k). Ac1 04×4 04×4 ∗ . (11) and ωi (i = 1. B∗ c Ed ˆ E1 = . 4×2 If the control input is limited by u(k) ≤ u0 .e. . 0 ˆ E2 = u1 (k) = Icmd. Bd = B∗ (C1 B∗ )−1 . 10. . JUNE 2007 where Fig. . then the following modiﬁed control input can be applied: ueq (k) u0 ueq (k) u(k) = ueq (k) for ueq (k) ≤ u0 for ueq (k) > u0 . If the tracking/disturbance poles to be considered are ±jω1 . . ω3 . u1 (k) = Icmd. the continuous-time servocompensator is ˙ η = Ac η + Bc evdq . ω1 = ω. After the dynamics (10) of the DSMC is included in (7). the control input (u1 ) can be given as ˆ u1 (k) = −KX(k) = −[K1 K2 ] X(k) η(k) (17) = −K1 X(k) − K2 η(k). Note that only the dominant 5th and 7th harmonics are chosen as the disturbance poles because the voltage harmonics such as an odd multiple of 3 and even harmonics are suppressed in a three-phase inverter. ω2 = 5ω. 2.). the current command signal (Icmd. −ω 2 · I2×2 02×2 4×4 i Bc3 12×2 Bci = 02×2 I2×2 . ˆ d(k) = ˆLdq (k). ˆ ˆ XT (k)QX(k) + εuT (k)u1 (k) 1 k=0 (16) For the above system (12). jε = ∗ Ad = A − B (C1 B ) C1 A∗ . As a result. Aci = . the overall plant can be expressed as ˆ X(k + 1) = Ad X(k) + Bd u1 (k) + Ed d(k). . ω3 = 7ω. ±jω2 . (13) where Q is a symmetrical positive-deﬁnite matrix and ε > 0 is a small number. evdq = VLdq − VLdq . . Also. NO. The optimal gain matrix K is obtained by solving the algebraic Riccati equation for (15). . η(k) Bd ˆ B= . yd (k) = Cd X(k) where yd = [VLdq ]. Therefore. 10.idq (k).472 IEEE TRANSACTIONS ON ENERGY CONVERSION. 3). Discrete-Time Optimal Voltage Controller As shown in Fig. B. (i. VOL. Discrete-time voltage controller. The discrete form of (13) is η(k + 1) = A∗ η(k) + B∗ evdq (k) c c where A∗ = eAc Tz ..
and a load change. Fig. and Ki is the integral gain. 11.54µs. and then duration (Tcal = Ta /3) of shootthrough zero vectors is calculated by look-up table according to the magnitude of the fuel cell output voltage. As ∗ illustrated in Table I. a = 0. and an error between the ∗ desired voltage (Vc2 ) and sensed voltage (VC2 ) is used as an input of the controller.3814. TABLE II SYSTEM PARAMETERS FOR SIMULATIONS Vc1 = Vc2 = Tb Tz − Ta 1 − Ta /Tz = = Vin . This controller should regulate the average dc-link output voltage to the desired value and the shoot-through duration (T ) should be conﬁned to a reasonable value that can guarantee accurately the required boosted dc-link voltage. Ta /Tz . Discrete PI dc-Link Voltage Controller Finally.5 kW: Ta /Tz = 0. 11. and simulation studies are performed under both linear and nonlinear loads. As shown in Fig. M.2668. 12. Equation (19) shows how Tcal is theoretically calculated assuming that a desired capacitor voltage (VC2 ) is 340 V and a switching period (Tz = 185.4639. the parameters K. one of the capacitor voltages (VC1 .JUNG AND KEYHANI: CONTROL OF A FUEL CELL BASED Z-SOURCE CONVERTER 473 Fig. The system parameters are given in Table II. VC2 ) is measured for the feedback control. Discrete PI controller to regulate the DC-link average voltage. Relationship between Vin . 12 shows relationship between Vin . T = 6. Moreover.1053. C.2 µs).2158.6186. Ta /Tz . Tb − Ta Tz − 2Ta 1 − 2Ta /Tz (19) discrete PI controller. 2). SIMULATION RESULTS To validate the effectiveness of the proposed model and control strategy. a linear/nonlinear load. the simulations are implemented under various conditions: a heavy/light load. T = 23. M = 0. In case of the heavy load (10 kW). the reference duration (T ∗ = Ta /3) of shoot-through zero states is directly added to or subtracted from the conventional switching patterns of SVPWM to boost the average dc-link voltage. K = 4. When Vin = 130 V and P = 10 kW: Ta /Tz = 0. a = 0. we assume that the output voltage of fuel cell is 130 V. the output voltage (Vin ) of the fuel cell is sensed. and T can be calculated as follow: 1). M = 0. and K. and K. The calculated shoot-through duration (Tcal ) is utilized in a saturation block in order to limit the ﬁnal output (T ) of the ∗ e(k) = Vc2 (k) − Vc2 (k) T (k) = T (k − 1) + Kp (e(k) − e(k − 1)) + Ki Tz (e(k) + e(k − 1)) 2 (18) . K = 1. while in case of the light Therefore Ta /Tz = Vc1 − Vin 1 ⇒K= 2Vc1 − Vin 1 − 2Ta /Tz M (= Tb /Tz ) = 1 − Ta /Tz ⇒ a = 3M/4 T = Ta /3 Fig.5µs. a. 11. That is. When Vin = 300 V and P = 0. and as we expect.8947. Ta /Tz and K decrease as the output voltage (Vin ) of the fuel cell increases.671. In this paper. VI. a simulation test bed using Matlab/Simulink is constructed for an ac 208 V (L–L)/60 Hz/10 kVA. The equation for the PI controller is given (18) where Kp is the proportional gain. a discrete PI voltage controller based on Tustin’s method for approximation to integration is used to regulate the average voltage of dc-link in Fig.
22.8 mF. 13. 2. Fig. Rs = 0. 13 shows simulation waveforms under the heavy and light load: (a) heavy load (130 V and 10 kW).5 kW.02 Ω and Cs = 4. Simulation results under a linear load. Fig.5 kW).5 kW) that of fuel cell is 300 V.2 mF to show a general dynamic response of the fuel cell . when the load decreases from 10 to 5 kW. JUNE 2007 Fig. and (b) light load (300 V and 0. 15 and 16 show the results under a load change: load increase and decrease. 4. load (0. whereas the nonlinear load is composed of a threephase inductor (2 mH). From Fig. we assume that the parameters of the reformer and stack are: Rr = 0. VOL. Fig. a dc-link capacitor (800 µF ). NO.474 IEEE TRANSACTIONS ON ENERGY CONVERSION. Fig. respectively. and a resistor (7 Ω). a three-phase diode bridge. Also. Simulation results when the load increases. In addition. 14 shows simulation results under the nonlinear load (130 V and 7 Ω). 14. On the other hand. (b) 300 V and 0. and the shoot-through in which both power switches in a leg are simultaneously turned on is deﬁnitely . The linear load consists of a resistor. 17 shows PWM waveforms of six-power transistors at heavy load to regulate the average dc-link voltage (VC2 ) as well as the inverter output voltage. (a) 130 V and 10 kW.05 Ω and Cr = 42. Simulation results under a nonlinear load. when the load increases from 5 to 10 kW. Figs. the output voltage of the fuel cell is changed from 130 to 250 V. 15. we assume that the output voltage of the fuel cell is changed from 250 to 130 V. Fig.
Estimated load current under a heavy load. Fig. vol. and load changes. (c) Load line to line voltages (VLAB . (d). A space vector pulse-width modulation technique is modiﬁed to realize the shoot-through zero vectors that boost the dc-link voltage. (d) Inverter output phase currents (iiA . 39. vol. To validate the proposed method. Keyhani. Marwali and A. iiB . First of all. VII.” IEEE Trans.. As shown in Figs. and a voltage–current polarization curve of a cell is also considered. iLC ). pp. Power Electron. Keyhani.. S. N.  A. 19. pp. and D. VLBC .  M. and low Total Harmonic Distortion (THD). “Reliability modeling of distributed generation in conventional distribution systems planning and analysis. Agarwal. “Control of distributed generation systems. 2003. and (e) Load phase currents (iLA . the efﬁciency can be improved in the range of several percent over the conventional dc/dc boost converter plus VSI because of no switching and conduction losses due to power switches of the dc/dc boost converter. Also. W. enhanced reliability. and lower cost at the expense of the added complexity in control. N. a zero steady-state voltage.–Oct. and A. it has some advantages such as increased efﬁciency. each ﬁgure indicates (a) Power request at 70 ms.” IEEE Trans. O. In Figs. 18. a system with a three-phase ac 208 V/60 Hz/10 kVA is simulated using Matlab/Simulink under various operating conditions such as heavy/light loads.. shown. 2004.” IEEE Trans. PWM waveforms at heavy load. 18 shows the real load current (iLA ) and estimated load current (i∗ )under the heavy load (130 V/10 kVA): upper LA (linear load) and lower (nonlinear load). 1493–1498. 1551–1561. linear/nonlinear loads. Ind. iLB . Nov. Nov. when comparing the Zsource converter with dc/dc boost converter plus voltage source inverter (VSI). (b) Output voltage of the fuel cell (Vin ) and capacitor (VC2 ). iiC ). Marwali. a good voltage regulation of capacitor voltage (VC2 ) is presented in spite of a slow change of the fuel cell output voltage and load changes. voltage/current sensors and a DSP controller as part of a dc/dc boost converter. sensors and DSP controller to boost dc-link voltage are removed. Appl. Part I: Voltages and currents control. 19. Power Electron. Fig. Jung. As a result. Sep. An R–C circuit model is used to realize slow dynamics caused by a chemical/electrical response of the reformer and stack of the fuel cells. REFERENCES  M. 2004. Simulation results when the load decreases. Chowdhury. Especially. . Fig. 15 and 16 (b). Part II: Load sharing control. a new control algorithm is presented which can guarantee a fast. Koval. J. vol. 13 to 16. no-overshoot current response. 1541–1550.JUNG AND KEYHANI: CONTROL OF A FUEL CELL BASED Z-SOURCE CONVERTER 475 Fig. A. the total system cost can be reduced by tens of percent because the power devices. modiﬁed space vector PWM implementation and design of a closed-loop controller of the Z-source converter employing L–C passive components. VLCA ). CONCLUSION The paper studies system modeling. “Control of distributed generation systems.Furthermore. 17. pp. K. the Z-source converter does not need power devices. 16.
 K. 1618–1623. power converters. G. M. Jun. N. 709–715.-H.  K. and M. 2. vol. OH. “The internal model principle for linear multivariable regulators. vol. A. Oct. Jin-Woo Jung (S’97–M’06) received the B. 2. and 2003 . 727–733. Enjeti. L. ac motor drives.. he is a Professor of Electrical Engineering at the Ohio State University. Jun. 442–450. vol.-H. 2667–2674. Campbell. 1999.. Pukrushpan. Lee. J.  R. 144–145. and J. Howze. no. M. 1999. pp. Y. “A modiﬁed dead-beat control for UPS using disturbance observers. pp. K. Samsung SDI Company. 309–314. Hauer. Control Conf. T. “Dynamic interaction between the electric drive train and fuel cell system for the case of an indirect methanol fuel cell vehicle.” IEEE PESC’02. Becker. pp.  B.476 IEEE TRANSACTIONS ON ENERGY CONVERSION. Feliachi. and F.” Electric Supply Industry in Transition. Gopinath. Chiu and B. Kjaer. vol. R.” IEEE PESC’02. A. Autom. and H. and J. Jan. 1999.. Control. H. 339–346. Lukas. Sliding Mode Control in Electromechanical Systems. “Z-source inverter. 22.. J. Soc. Jun. 504–510. Jun. 1996. Ponci. vol. vol. Burghardt.  L. Korea. vol. no. 2004. 689–702. Lukas. A. vol. Energy Convers. Summer Meet. 4. Soc. Franzoni. D. 28–36. and the Ph. pp. 14. vol. control of power systems.” Proc. G.” IEEE Trans. Energy Convers. 1. F. Ghezel-Ayagh. 170–194. and H. A. Pinheiro.” IEEE Power Eng.. vol.  M. Grundling. Andersen. Pinheiro. 3. He is the Director of Ohio State University Electromechanical and Mechatronic Systems Laboratory. Kim and S. pp.. Energy Convers.-W. 2. pp. “Dead beat microprocessor control of PWM inverter for sinusoidal output waveform synthesis. and H. Gokhale..  J. “An improved small-signal model of the dynamic behavior of PEM fuel cells.” Am. Lee. Dr. LG Electronics Company. pp.L. 2003. AIT Thailand. pp. Currently. 2002. and R. pp.” IEEE Trans. Seoul. Spec. Yeary.  L. degree in electrical and computer engineering from The Ohio State University.” IEEE Trans.” 35th IECEC Meet. Ghezel-Ayagh. Dec. Webster. Jan. 14.” IEEE IAS’01.. T. Philadelphia: Taylor & Franci. 8. Stefanopoulou. Monti. Ltd. Jul. and failure detection systems.S. 1663–1665.S. Blaabjerg.. and A. Del Monaco. pp. “A low-cost inverter for domestic fuel cell applications. H. S.. Opt. Jung and A. “Digital voltage and current controllers for three-phase PWM inverter for UPS applications. of ASME IMECE’02. 2. Columbus. Guldner. Kim. Utkin. Currently.  F. 14.” IEEE Trans. “An explicit dynamic model for direct reforming carbonate fuel cell stack.. Z. “Design of Z-source converter for fuel cells. 289–295. S. D. Power Electronics. 1999. in 2005. vol. 2. Hey. 2002. “A fuel cell based domestic uninterruptible power supply. Sep. 2000. Kawamura. “Deadbeat control of a three-phase inverter with an output LC ﬁlter. Ali Keyhani (S’69–M’72–S’72–M’76–SM’89– F’98) was with Hewlett-Packard Company and TRW Control from 1967 to 1972. Hahn. he was with Digital Appliance Research Laboratory. Pukrushpan. Peng. Jun. Klumpner. Peng.  Y.  M. 11. “Modeling and control for PEM fuel cell stack system. pp. 2. VOL. pp.  O. “Simulation and analysis of transient fuel cell system performance based on a dynamic reactant ﬂow model. and A. vol. Barry. control and design of power electronic systems. His current research interests include the control of distributed energy systems. Ind. no. pp. NO. vol. vol. Patterson. Winter Meet. M. automotive systems. “A new green power inverter for fuel cells.–Apr.  J. B. and N. Scherzinger.  F. M. Summer Meet. Gyeonggi-Do.D. “Use of renewable sources of energy in Mexico. “Dynamic modeling and analysis of polymer electrolyte fuel cell. vol... pp. Krase. electric machines. Conf.  J. “Development of a stack simulation model for control study on direct reforming molten carbonate fuel cell power plant. 1975. Keyhani. Seoul. 2. Ltd. 82–86. 1. pp. Mar. H. 4. He is the past Chairman of Electric Machinery Committee of IEEE Power Engineering Society and the past Editor of IEEE TRANSACTIONS ON ENERGY CONVERSION. vol. H.. 1987. His current research interests are in the area of distributed generation systems. 2001. Mattavelli. Diong. A. Sep. Energy Convers. design and modeling of electric machine.” IEEE PESC’02. Peng.” IEEE Power Eng. Wonham. Appl. “The role of distributed generation in the critical electric power infrastructure. K. P. 239–244. 1. 2002. 2002. Record IEEE Power Elec. OH. “Development of a low cost fuel cell inverter system with DSP control.” IEEE APEC’02.  P. degrees in electrical engineering from Hanyang University. 2002. fuel cell systems.” IEEE Trans.” IEEE PESC’02. M.” IEEE Trans. Gutierrez-Vera. Botteron. 605–613. P. Colclaser. respectively. J. He is with Corporate R&D Center. pp..  D.” Appl. J. pp. Philipson.  A. Korea.  G. Energy Convers. Kukrer. JUNE 2007  J. 39. Jul. pp. “Distributed and dispersed generation: Addressing the spectrum of consumer needs. Hoft. vol. 2. Y. pp. Stefanopoulou. Davison and B. D. and H. 2000. S. 2003. 1317–1325. 2001. 1.” Conf. Francis and W. G. Sep. Soc. Keyhani is a recipient of the Ohio State University College of Engineering Research Award for 1989.” IEEE Trans. Yerramalla. 32.” IEEE Trans. D. Mar.  J. Columbus. pp. no. Santi. 1. Math.  E. pp. Meet. parameter estimation. 2002. Y. 9. K. 3117–3122. J. 1994. modeling. A. Shi. From 1997 to 2000. and control. C.-S.–Feb.  V. 2002. J. vol. Korea in 1991 and 1997. Kim. driving circuits of ac plasma display panel (PDP). G. pp. vol.. Tuckey and J. “An electrical modeling and fuzzy logic control of a fuel cell generation system. 2001. vol. 16. 16–23. Davari. D. 4. 1985.  S. vol. “Perfect control of the robust servomechanism problem.” IEEE 38th IAS Ann. DSP-based virtual test bed for design. 1651– 1657. 749–753.” IEEE Power Eng. Jul. 1999. May 2002. Hall and R..  E. “Transient modeling and simulation of a tubular solid oxide fuel cell. Yongin.
This action might not be possible to undo. Are you sure you want to continue?