This action might not be possible to undo. Are you sure you want to continue?
Remus Teodorescu and Frede Blaabjerg
Institute of Energy Technology Aalborg University Aalborg, Denmark email@example.com, firstname.lastname@example.org
Marco Liserre and Antonio Dell’Aquila
Department of Electrotechnical and Electronic Polytechnic of Bari Bari, Italy email@example.com, firstname.lastname@example.org value for the ac inductance can solve this problem, but this makes it expensive and bulky. On the contrary, to adopt an LCL- filter configuration allows to use reduced values of the inductances (preserving dynamic) and to reduce the switching frequency pollution emitted in the grid . The main goal is to ensure a reduction of the switching frequency ripple at a reasonable cost and, at the same time, to obtain a high performance rectifier. Usually the converter side reactor is bigger than the grid side one because it is responsible for the attenuation of most of the switching ripple. The ac capacitor is limited in order not to reduce too much the power factor and the grid side reactor is chosen in order to properly tune the cut-off frequency of the LCL-filter. The stability of the system should be rigorously studied . A poor analysis made on qualitative considerations could lead to excessive damping (unnecessary increase of losses) or insufficient damping (the system seems to be stable but it is not). The damping, both passive (based on the use of resistors) and active (based on the modification of the control algorithm) should be rigorously tested . In this paper the design of an LCL-filter based active rectifier is reported which is stable even without damping due to a proper choice of the passive elements of the LCL-filter and to the position of the current sensors on the grid side rather than on the converter side. The reported experimental results prove the theoretical analysis. II. SYSTEM CONFIGURATION The equivalent single-phase circuit for the three-phase active rectifier system is shown in Fig.1. The following base values will be introduced:
Abstract—Industrial LCL-based grid converters using an active
rectifier need to be designed in view of robustness, stability and high efficiency. In this paper the design of an active rectifier that does not need damping whether passive or active is described. This allows obtaining stability without the decrease of the efficiency typical for the passive damping methods or the increase of cost due to more sensors or more complex control algorithms that is typical for the active damping methods. This has been achieved with a careful choice of sensor position and of the passive elements in the LCL-filter. In fact if the current sensors are on the grid side, rather than on the converter side, the current loop is much more near to the stability. Moreover, if the grid side inductance is a fraction of the converter side one, the current loop is again much more near to the stability. Thus, with a proper design the system can be made stable at some switching frequencies even without any damping. Keywords—Voltage Source Converter, active rectifier, digital control, harmonics
I. INTRODUCTION The most common applications for dc power systems are chemical, electrolysis, aluminum, graphitizing furnace, zinc electrolysis, copper refining, traction substation, AC and DC drive systems. The dc voltage regulations is often required and realised with traditional methods such as saturable core reactors and on-load tap changers or with phase controlled rectifiers, or finally with a chopper (dc/dc converter) on the dc side. All these solutions can hardly comply with power quality standards and often lead to overrating of the overall system. Moreover, the most advanced of them, the chopper, increases the number of switching devices of the system resulting in a higher failure rate and mean time to repair. On the contrary an active rectifier based on a Voltage Source Converter (VSC) with its ac side connected to the grid and employing pulse width modulation (PWM) has the capability to control the grid current on the ac side and the voltage on the dc side . In this configuration the grid current is synchronized in order to be in phase with the grid voltage and in order to obtain a unity power factor. The switching frequency is generally between 2 kHz and 15 kHz and causes high order harmonics that can disturb other EMI sensitive loads/equipment on the grid. Choosing a high
Z b = ( En ) Pn
(1) (2) (3)
Cb = 1 ω n Z b
ωb = ω n
where En is the line to line rms voltage, ωn is the grid frequency and Pn is the active power absorbed by the converter in rated conditions. Thus the reactances of L ( X ), of Lg ( Xg ) and of Cf ( XC ) are expressed in [pu] of the base impedance Zb: x = X Z b , xg = X g Z b and xC = − Z b X C . ωb is the base frequency considered equal to the grid frequency.
0-7803-7883-0/03/$17.00 © 2003 IEEE
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on February 9, 2009 at 06:48 from IEEE Xplore. Restrictions apply.
iron core.B. as shown by the Fig. where ω = 2πf and f is the fundamental frequency of the By considering a dq-frame that rotates at the angular speed grid. The specifications of the power stage of the VLT5022 frequency inverter from Danfoss are: Uout = 3x400V. they have the same plant. Irated = 16 A@50Hz with a ripple of 5 App@8kHz. Usually the damping of the system is chosen in a qualitative manner. Restrictions apply.B. Downloaded on February 9. Further details can be found in . Cf = L Z b2 2 2 2 where ω res = LT z LC L and z L C = L g C f [ ] −1 (5) The designed LCL-filter  is reported in the followings. Design guidelines for both passive and active damping are given in . Grid side inductance: Lg = 1.5 A.max = 850 V. Current-controlled three-phase active rectifier. resulting in a decrease of efficiency and unknown dynamic of the system. Irated = 16 A@50Hz with a ripple of 2. Capacitor Cf = 6 µF. Space Vector Modulation (SVM) is used as modulation strategy and a Phase-Lock Loop (PLL) is used for grid synchronization. 3a) for current control is (4) Hence in order not to overrate the VSC the following formula can be used to have an idea of the value of the ac side capacitor: G (s) = ig ( s ) v( s ) = 1 1 2 LLg C f s ( s 2 + ω res ) (6) . .2 The control structure of the active rectifier 1553 Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. 1 The following equivalent impedances are considered for the grid side and for the converter side. because the dc voltage controller calculates the reference value for the d-axis current controller (Fig. 2).3b when a damping resistor is inserted in series with the filter capacitor: G (s) = ig ( s ) v( s) = ( C f Rd s + 1) 1 2 2 2 LLg C f s ( s + ω res C f Rd s + ω res ) (7) U DC* U DC controller id* id controller ∆e d ωL iq controller ∆e q -ωL ed U DC GRID D A CD i q =0 * eq ed* s a.rated = 37.C e A. 1.  and . The model of the current loop shown in Fig. respectively: ω.Lg Cf L L III. Stability analysis If the grid side current is sensed the plant (Fig. thus the part of the system to the left of the capacitor works like a resistor. - with different approaches.s b . Iout. xTgrid = 1 xTconv = 1 + j ( x − xC ) A. CONTROL OF THE ACTIVE RECTIFIER VSC e ig CURRENT CONTROL ≡ Zb Cf (a) (b) Fig.C ac current controller θ PLL eX eY 23 Fig. stability performance and fast response. a) Single-phase diagram. ferrite core. Cnverter side inductance: L = 2 mH.and q-current loops are equivalent in the stability and dynamic analysis because once done the compensations of the grid voltage ed and eq and due to the cross-coupling ωL id(t) and ω L iq(t). Thus the controller structure is cascaded. UDC. 2009 at 06:48 from IEEE Xplore. . In the system the grid voltage is sensed and the grid current is controlled to be in phase with that.s c e q* id i q dq ab SVM D A DA VLT 5022 POWER STAGE LCL i A. Usually the CC is implemented in a dq-rotating frame that has the d-axis oriented on the grid voltage.5 App@8kHz. the d.5 mH. b) Equivalent circuit The ac current control (CC) is considered particularly suitable for active rectifiers for its safety.
This is an opposite situation in respect to that one obtained when the current sensors are on the converter side. 4d demonstrate that if there is one delay in the current loop due to the use of filters or due to too long computation times the system is less stable and passive damping is needed. It is interesting to note that the system is stable even without the damping (Fig 4a). Then Fig. instead the use of the damping (4 Ω in series with the filter capacitor) makes the system to be more stable but moves the complex conjugate poles on the right half of the plane away from the optimum position (Fig.3. Downloaded on February 9. 2009 at 06:48 from IEEE Xplore. Single-phase equivalent of the current loop plant undamped (a) and damped (b). too. 4. In that case one delay in the current loop makes the system more stable and reduce the need for damping . The results of the stability analysis made at a switching frequency of 8 kHz are reported in Fig. A.v _ + 1 sL i ig e Lg Cf ic L vc i v vc 1 sC f ig _ + e (a) _ + 1 sLg v _ + 1 sL i ig e Lg ic L vc i v Cf vc Rd e (b) 1 + Rd sC f ic _ + ig _ + 1 sLg Fig. Design algorithm Once the current control loop stability has been analysed.4b). Moreover the maximum proportional gain (kmax) for which the system is stable is highlighted. . 4c and Fig. The first step is to choose the parameters to be optimised. Instead the damping losses related to the presence of Rd in series with the LCL-filter capacitor Cf can be estimated considering the transfer function of the capacitor current ratio the converter voltage G2 ( s ) = ic ( s ) s 1 =− 2 2 v( s) L ( s 2 + ω res C f Rd s + ω res ) (9) again evaluating G2(s) for s=jhω and for s=jhωswitching because Cf drains a considerable current only at the fundamental frequency and for multiple of the switching frequency. They have been defined also with the help of two transfer functions in the s-domain. 1554 Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Restrictions apply. The transfer function of the grid current ratio the converter current is used to calculate the LCL-filter effectiveness in reducing the switching ripple G1 ( s ) = ig ( s ) i(s) = (L C g (C f s + C f Rd s + 1) 2 f Rd s + 1) (8) evaluating G1(s) for s=jhωswitching where ωswitching=2π fswitching and h is an integer because the adopted modulation generates harmonics around the multiples of the switching frequency. In the figures there are the root loci with highlighted positions of the roots when the proportional gain of the current controller is chosen on the basis of the technical optimum (koptimum) . a design procedure can be developed in order to minimise the damping for the LCL-filter based active rectifier.
5 (b) 0 0.1 k max k optimum k max k optimum 1 k max k optimum k max k optimum 0. Root locus of LCL-based active rectifier at 8 kHz sampling and switching frequency.ρMAX should be at least < 1 in order to have a stable current loop. for the LCL-filter switching ripple attenuation: ra – it can be calculated on the basis of G2(s). Downloaded on February 9. (c) the undamped current loop with one sample delay and (d) the passive damped current loop with one sample delay Thus the parameters used to assign the design constraints are 1.5 1 -1 -0. (a) the undamped current loop. 4.maximum radius of poles of the current closed loop .5 0.5 1 -1 -0. The current controller proportional gain kp.5 0. 3. for the stability ρMAX .5 -1 -1 -0. 3.5 (c) -1 0 0. Three parameters have been chosen to be adapted in order to find the optimum solution: 1.5 0 0 -0.5 -0.5 -0. 2. for the bandwidth: bw – the lowest between the frequencies at which the gain of the closed loop is reduced to 3 dB and at which the phase delay becomes larger than 45.5 0 0.5 0 0 -0.5 1 Fig. The damping resistor value Rd. 1555 Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. 2. 2009 at 06:48 from IEEE Xplore.5 1 (a) k optimum 1 k optimum 1 k optimum k max k optimum k max 0. The sampling frequency fsampling. 4. for the damping losses: Pd – they can be calculated using G1(s) since they depend on the square of the LCL-filter capacitor Cf rms current Ic. Restrictions apply.5 -1 -1 -1 -0. (b) the passive damped current loop.5 (d) 0 0. .
Pd depends on the last two of them in a nonlinear way and ra depends especially on the second of them.Roughly speaking ρMAX is a function of all the three parameters but especially of the last two in a non-linear way. The test conditions were: UDC = 650V. Pout=11. phase voltage ea and current ia. Rd) and Pd(kp. in this case 650 V. fsampling. 1556 Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. Downloaded on February 9. Fig. Current probe is reversed as in rectifying mode phase voltage is in phase with the phase current. Steady state operation The steady state operation is shown in Fig 5. Rd). THDIgrid = 6. 7b full-load transient is applied from no-load condition. The method gives good results if the optimal solution is near the initial conditions. using the optimisation method for designing the LCL filter. ra(kp. As it can be observed in Fig.7 %. THDUgrid = 2. It has been chosen to use the Levenberg-Marquardt method  in conjunction with the linear search . Otherwise. Finally. The transient response is good showing a settling time in the range of 30 ms. Many parameters can be settled in the “lsqnonlin” algorithm such as minimum and maximum change in variables for finite differencing and the method of interpolation. the dc voltage increases to the reference value. The minimum along the line formed from this search direction is approximated by a polynomial method involving interpolation.0 %. EXPERIMENTAL RESULTS In the following the steady state and the transient operations of the converter are shown. bw(kp. prior to starting the rectifier mode the dc-link is charged. The results demonstrate that at a switching frequency of 8 kHz the system is stable even without any kind of damping. Using this optimization method the present LCL filter has been designed. IV.2 kW. UDC C ea ia (a) UDC C ea ia (b) Fig. PF = 0. at the peak value of the grid voltage. After starting the rectifying mode. A. 2009 at 06:48 from IEEE Xplore. 5 Phase voltage and current at nominal load (11kW) in rectifying mode (a). The line search is based on the solution of a subproblem to yield the search direction in which the solution is estimated to lie. Transient operation The system start-up in rectifying mode is shown at noload condition and full-load condition.99. Ugrid. fsampling. The Levenberg-Marquardt method uses a search direction which is a solution of a linear set of equations. Thus a step-by-step algorithm can be written in order to tune the parameters simply exploiting the previously reported dependencies. Rd). the grid current looks stable in stationary conditions at rated load. Steady state operation was measured using the oscilloscope Tektronix TDS 3014 and power analyzer PM3000A. Rd). Polynomial methods approximate a number of points with a polynomial whose minimum can be calculated easily. The current peak at startup is in the range of 150% of the rated current but this is due primarily to the fact that the current control was not active prior to the transient. As it can be observed. . The dc-voltage restoration time is again in the range of 30 ms but now the overcurrent has disappeared as the current control was active before the transient. fsampling. In fact this method has proved to be more robust than the GaussNewton method and iteratively more efficient than an unconstrained method . bw depends strongly and almost linearly on the second parameter. 5. 7a a 4 kW load transient and in Fig. a non-linear least-meansquare method can be adopted in order to find the optimal solution without linearising the relations ρMAX(kp. generating mode (b) B. 6 shows the DC-link voltage UDC. as shown in II. The difference in the appearance of the current is due to the fact that in generating mode the fundamental of the current is 180 degrees shifted and so the superimposing of the low harmonics from the grid voltage leads to different results. in Fig. through the freewheeling diodes from the converter IGBT bridge. Restrictions apply.phase = 230V. fsampling. The optimisation method has been tested through the Matlab® algorithm “lsqnonlin” .
guarantee stability and maximize the LCL-filter attenuation. pp. pp. Liserre. Springer Verlag. B. of APEC 2003. 2001. “A control method to damp oscillation in the input LC filter of AC-DC PWM converters. Twining. Università e Ricerca" (project: CLUSTER 13).” Proc.  V. Downloaded on February 9. Kaura. 1977. pp. App. Slemon. and F. “Genetic algorithm based design of the active damping for a LCL-filter three-phase active rectifier. Liserre. 27.  P. on Ind. March/April 1991. nominal load (11kW) (b) in rectifying mode In all cases no sign of instability in the current could be observed. 542-550. 6 Start of rectifying mode at no load (a). Numerical Analysis. 7 Step load change from no load to 4kW load (a).  E.  R. Moreover. J./Oct. Restrictions apply. Dell’Aquila. Lecture Notes in Mathematics 630. vol. of IAS 2001. of PESC 2003. June 2002.  M.J. Proc.  Moré. A.Blaabjerg. of PESC 2002. . and F. R. 355-364. F. and S. pp. using the optimal design method V. Watson. Sept. “A novel control to actively damp resonance in input lc filter of a three-phase voltage source converter. A. G. Dell’Aquila. demonstrating thus that the system is stable even without any damping at all. Blaabjerg. Liserre. Dewan.UDC C UDC ea ia ea ia (a) (a) ea UDC ia C UD iDC ia ea (b) Fig. pp. June 2003. on Ind. ed. F. full load (b) (b) Fig.” Proc. G. June 2002. a design method has been proposed to minimize losses. Vol. Blaabjerg. “ Stability improvements of an LCL-filter based three-phase active rectifier. February 2003. REFERENCES  R.” IEEE Trans. F. pp. Dahono.” Proc. and D. 16301635. July 2002 1557 Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE MINAS GERAIS. App. 2009 at 06:48 from IEEE Xplore. pp 105-116. 1189-1194. 1195-1201.  Matlab Optimization Toolbox User’s Guide. “Grid current regulation of a threephase voltage source inverter with an LCL input filter.” Proc. Wu. and G. CONCLUSIONS The paper deals with the design of a 11 kW active rectifier with LCL-filter designed to be stable even without damping avoiding unnecessary losses and the decrease of efficiency. A.Iov. of PESC 2002. ACKNOWLEDGEMENTS This work has been carried out with the financial support of the Italian "Ministero della Istruzione.  M. The Levenberg-Marquardt Algorithm: Implementation and Theory. increase bandwidth. Hansen. “Analysis of an ac-to-dc voltage source converter using PWM with phase and amplitude control. “Flexible test and development system for a 11 kW wind turbine”.” IEEE Trans. 33..” Proc. S.  M.. and V. Blaabjerg. Holmes.Teodorescu. “Design and control of an LCL-filter based active rectifier. March-April 1997. 299-307. Blasko. of PESC 2002. A. June 2002. Experimental results in both stationary and transient conditions demonstrate the success of the new design method..
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.