This action might not be possible to undo. Are you sure you want to continue?
Study of Advanced Current Control Strategies for Three-Phase Grid-Connected PWM Inverters for Distributed Generation
Qingrong Zeng and Liuchen Chang, IEEE Senior Member
Abstract — Three-phase PWM voltage-source inverters (VSI) are widely employed for distributed generation (DG) systems. Since the voltage at the point of common coupling should not be regulated by DGs, the current control strategy of gridconnected inverters plays a dominant role in providing high quality power to electric grids. The Authors have proposed two advanced SVPWM-based current controllers to improve the performance of grid-connected VSIs for DG systems. Three current control strategies, namely hysteresis current control, SVPWM-based PI control and SVPWM-based predictive control, have been developed, implemented, and studied on a 30kW three-phase grid-connected PWM VSI. Effective compensation for the grid harmonics and for the system nonlinearity due to control delay and switching dead time are investigated. The results of computer simulations and experimental tests are provided to verify and compare the performances of these three current control strategies.
Global demand for electricity is estimated to increase from 13,290 TWh in 2001 to 23,072 TWh in 2025 . With the Kyoto Accord aiming at reduction in greenhouse gas emissions, fossil-fuel based power generation alone can no longer meet the demand for new generation capacity , which presents an unprecedented opportunity for distributed generation (DG) systems. Most DG systems need power electronic converters, often referred as inverters, to realize power conversion, grid interconnection and control optimization. Pulse-width-modulated (PWM) voltage source inverters (VSI) are widely applied in DG systems. Inverters are critical in DG systems for optimizing the control functions as well as meeting the interconnection and power quality requirements . The current control of a grid-connected VSI determines the power quality with which a DG feeds a grid. Many current control strategies have been developed for three-phase PWM VSIs -. In most three-phase PWM VSIs for DG systems, ac motor drives, active filters and uninterruptible power supplies, current control strategies have a similar structure with an inner current feedback loop, accomplishing two basic tasks: the current error compensation and PWM generation . Current control strategies can be classified into two major categories: nonlinear controllers based on close-loop current-type PWM and linear controllers based on open-loop voltage-type PWM . As a nonlinear controller, hysteresis current control (HCC) is widely employed for three-phase PWM VSIs -. With HCC, current errors are directly used to determine the next PWM state of the VSI. The current error compensation and
The Research is supported by Natural Sciences and Engineering Research Council of Canada (NSERC). The Authors are with Dept. of Elec. & Comp. Eng., University of New Brunswick, Fredericton, NB, Canada E3B 5A3. (Phone: 506-447-3145, E-mail: LChang@unb.ca.)
PWM generation are carried out in the same control unit at the same time, rendering the controller a good dynamic response and an inherent current protection . However, high current ripples and variable switching frequency usually result in poor current quality and difficulties in output filter design. Among linear controllers, current control strategies based on space vector PWM (SVPWM) are widely employed . A SVWPM-based current controller separates current error compensation and PWM functions, making it possible to exploit the advantages of SVWPM as well as to independently design the overall control structure . SVPWM has many advantages such as constant switching frequency, well defined output harmonic spectrum, optimum switching patterns, and excellent dc-link voltage utilization . However, as a voltage-type modulator, SVPWM has a compromised output current quality due to the back-EMF disturbance of loads/grids and nonlinearity of systems, and lack of inherent over-current protection . To overcome these drawbacks, the design of current error compensation becomes critically important. Among the previously developed linear current controllers, proportional-integral (PI) regulation and predictive control are two dominant algorithms for current error compensation. Conventional PI regulators normally do not have appropriate compensation for the grid harmonics , yielding poor output current waveforms for grid-connected applications. Predictive control algorithms rely on a deadbeat control for VSI current control, however, is quite complicated and sensitive to system parameters . Operating in stationary coordinates, a previously developed direct digital predictive current control strategy  suffers from a degraded performance due to control delays. Two advanced and robust SVPWM-based current controllers were developed by the Authors  on the synchronous rotating d-q reference frame. In this paper, three current control strategies, i.e. HCC, SVPWM-based PI control and SVPWM-based predictive control, are studied on a 30kW three-phase grid-connected PWM VSI. II. SYSTEM DESCRIPTION
As shown in Fig. 1, the power conversion system for a 30kW three-phase grid-connected DG system has a typical ac-dc-ac topology with a dc-link boost chopper. A three-phase diode rectifier converts the variable input ac voltage into dc voltage. A basic boost chopper maintains the dc link voltage at an appropriate level to ensure a high quality current feeding into the grid even at a low ac input voltage. A typical three-phase six-IGBT full-bridge PWM VSI is connected to the grid through an inductor filter. The control platform of the entire system is built on a TI DSP microcontroller of TMS320LF2407A .
0-7803-9354-6/05/$20.00 ©2005 IEEE
Fig. however. vd and vq are output voltages in d-axis and q-axis. 1 Block diagram of a gird-connected power conversion system. The SVPWM-based PI current controller is constructed based on (1). 4 depicts the current error compensation block as part of the developed PI current controller of Fig. with different algorithms for current error compensation. In the decoupled d-axis and q-axis current control loops. The resulting three current errors are directly used to generate the PWM signals for IGBTs through hysteresis comparators. T2 Vdc Idc IGBT DRIVE BOARD Ia Ib PWM 6 GRID VOLTAGE SENSOR Vga Vgb Vgc INTERFACING BOARD TMS320LF2407A DSP BOARD A SVPWM-based linear controller in the synchronous rotating frame can offer an appropriate current control strategy with a high quality output in three-phase gridconnected VSIs. Two advanced SVPWM-based linear current control strategies with a PI controller and a predictive controller have been developed by the Authors. 2. III. the SVPWM-based PI control strategy is presented. HCC generates the desired PWM signals for the next control cycle so as to maintain the current trajectories within the hysteresis bands . HYSTERESIS CURRENT CONTROL A basic hysteresis current control was developed and implemented in the VSI. 3 Block diagram of the proposed SVPWM-based current control system for grid-connected VSIs. Control System Configuration To improve the current control performance. HCC is insensitive to system iA* + iB * vd = L ⋅ did − ω ⋅ L ⋅ iq + v gd dt diq + ω ⋅ L ⋅ id + v gq vq = L ⋅ dt (1) iA + ia ib iB + ic iC PWMA 2 PWMB 2 PWMC 2 3-Phase VSI vdc iC* Fig. A. L Grid parameters and is extremely simple for implementation. Fig. which iq virtually eliminates the ˆ iaF ability to generate zero ˆ iq ia Grid ˆ ibF Predictor voltage vectors. Three-phase output currents of the inverter are detected and compared with the corresponding phase current references individually. 2 Block diagram of hysteresis current control strategy for a three-phase grid-connected PWM VSI. The outputs of PI controllers are inductor filter voltage references VLd* and VLq* that are superimposed by Vgd and Vgq to generate the inverter output voltage references Vd* and Vq* for SVPWM. Moreover. as shown in Fig. and are studied in this paper. is grid angular frequency. as proposed by the Authors. L is output filter inductance. 3. where id and iq are output currents in d-axis and q-axis. due to the closed-loop PWM. 3. The grid voltage vector is used in a feed-forward loop to compensate for the grid harmonics. vdc Grid However. since the L vdc * * three-phase currents are id PWM vd 3-Phase Current Error independently controlled * * SVPWM 6 vq VSI Compensation with a control delay.AC INPUT RECTIFIER BOOST CHOPPER INVERTER LPA POWER GRID IV. In this section. They have the same control structure as shown in Fig. if the back-EMF harmonics and system nonlinearity due to control delays and switching dead time are well compensated. the SVPWMbased PI current controller is developed in d-q synchronous reference frame where all ac variables become dc variables. vgd and vgq are grid voltages in d-axis and q-axis. Based on Park Transformation . In both the current feedback loop 1312 . Clearly. HCC offers an inherent current protection and usually an outstanding dynamic response. two conventional PI controllers are employed to eliminate current errors. the output 2r ← 3s ˆd ib i Phase-voltage ˆ & Filter current ripples may be icF Detection quite large and the total vga Grid Angle harmonic distortion (THD) vgb Detection of the output currents vgc ˆ v gq vgq could be unacceptably Predictor v ˆ v gd gd 2r ← 3s high for power grids. the voltage equations of a three-phase grid-connected VSI in the synchronous d-q reference frame can be described as : Fig. SVPWM-BASED PI CURRENT CONTROL V V Vgen HEAT SINK LPB LPC 6 IGBT DRIVE BOARD Vgen PWM Iboost T1.
where id(n) and iq(n) are the actual direct and quadrature currents at the end of the present PWM cycle. the sampling frequency is set at 4 times the PWM frequency. a software predictor and filter are employed to estimate the output current and grid voltage for improved steady state and dynamic performance. 4. and D in Fig.q (K − 3) V. assuming that the grid voltage is linear during one PWM cycle. A digital predictive algorithm has been developed by the Authors as the current error compensation block for calculating the output voltage references of VSIs so as to eliminate the current error. 1313 . Ts is the PWM period. Then. 5. in a similar way as described in Section IV-B for PI current control. 3 1 ˆ ˆ i F ( K + 1) = ⋅ i ( K + 1) + i ( K ) 4 4 + + + - (3) iq*(n) ˆ id ( n) Lm / .q (K ) − vgd. 6. In each PWM cycle. 5 PWM carrier and sampling timing. and the sampling timing is synchronized with the initial point A of each PWM cycle. PWM carrier period (n+1)th cycle A (K-3) B (K-2) C (K-1) D (K) A (K+1) B (K+2) C (K+3) D (K+4) A (K+5) Fig. the grid voltage at (K+3) can be approximated as the average grid voltage in the next PWM cycle. As shown in Fig. a software RC filter is constructed. and the current control algorithm is executed only when the A/D conversion is finished at point D in order to minimize the control delay. B. The block diagram for this current error compensation is shown in Fig. the instantaneous values of id. With the symmetrical SVPWM. This sampling scheme requires a high-speed ADC and well designed control codes to guarantee the timely completion of the control algorithm.q(n) and vgd. Again. To eliminate the effect of the control delays due to sampling and computation. an A/D converter (ADC) is activated 4 times as indicated by A. to achieve output current references in each PWM cycle. In order to eliminate the effect of phase angle difference between K and (K+3).e.Lm / Lm /Ts + ˆ av v gd (n + 1) + vq*(n) The average grid voltage in the next PWM cycle is used to generate inverter output voltage references. the instantaneous values of output currents at the end of the current PWM cycle and the average values of grid voltages of the next PWM cycle are predicted. the prediction is done in the d-q reference frame. 4 Current error compensation of SVPWM-based PI current controller. * v d ( n) = L m ⋅ * * i q ( n ) + i q ( n) i d ( n) − i d ( n) av − ωLm + v gd (n + 1) Ts 2 * i q ( n) − i q ( n) i * ( n) + i d ( n) * av v q ( n) = L m ⋅ + ωLm d + v gq (n + 1) Ts 2 ˆ i ( K + 1) = 2 ⋅ i ( K − 1) − i ( K − 3) id*(n) (2) + - Lm/Ts + + vd*(n) In order to alleviate the control deviation resulted from the integral part of current error regulators due to the noise and nonlinearity of the system. 5. According to (1). the output current of VSI can be considered approximately linear from (K-3) to (K+1) during each PWM cycle. Lm is the modeled inductance of the output filter and should be close to the actual inductance L. i. the current value at (K+1) can be readily predicted from the sampling values at (K-3) and (K-1). the required output voltages vd*(n) and vq*(n) can be predicted by (6). B. Sampling and Prediction In this paper.and grid voltage feed-forward loop. C.q (K + 3) = 2 ⋅ vgd. 5. vgdav(n+1) and vgdav(n+1) are average values of grid voltages in the next PWM cycle. 6 Current error compensation of the predictive current controller. to achieve the direct and quadrature current references id*(n) and iq*(n) at the end of the next PWM cycle. at (K+3) can be predicted by (4). a dual-timer sampling scheme is employed . As shown in Fig. The grid voltage PWM carrier period nth cycle control delay ˆ iq (n) ˆ av v gq (n + 1) (6) Fig. (4) SVPWM-BASED PREDICTIVE CURRENT CONTROL ˆ id ˆ v gd Fig. SVPWM-based predictive current control strategy has the same control structure in Fig. id * iq * + + id iq ˆ iq PI PI vLd* vLq * + + ˆ v gq vd* vq* ˆ vgd.qav(n+1) in (6) have to be predicted accurately using the available sampling data.
the nominal line-to-line voltage of the three-phase grid was reduced to 208V/60Hz. 10 Simulation results of the dynamic response of SVPWM-based predictive control. the switching IGBT frequency was fixed at 9. For the hysteresis current control. the output filter inductance was 2mH per phase.02. 7 shows the simulated steady-state responses.766kHz. the two advanced SVPWM-based current controllers have much better steady state responses than that of HCC. the fundamental component of the grid phase voltage is 170sin(120πt).VI. the sampling frequency is 40kHz thus the control delay is 25uS. With the two SVPWM-based current controllers.3A. (c) inverter output phase-A current controlled by SVPWM-based PI controller (A). Fig. In all these cases. the compensation for the switching dead time is embedded in the SVPWM algorithms. 8 to 10 depict the simulated dynamic responses of the three current controllers. both HCC and the SVPWM-based predictive current control have fast dynamic responses while the SVPWM-based PI current control is inferior to the other two.3A. Lower: step change from 20A rms to 5A rms. VII. superposed by a 5th harmonic component 3sin(600πt) and a 7th harmonic component 8sin(840πt).0 and Fig. SIMULATION RESULTS 0. (d) inverter output phase-A current controlled by SVPWM-based predictive controller (A). Fig. EXPERIMENTAL RESULTS Fig. Upper: step change from 5A rms to 20A rms.0uS. (b) inverter output phase-A current controlled by HCC (A). Upper: step change from 5A rms to 20A rms. the switching dead time compensation was not needed for HCC. the hysteresis band was set at 0. Three current controllers have been implemented and tested on a 30kW three-phase grid-connected VSI. For the two advanced SVPWM-based current controllers. With HCC. the maximum IGBT switching frequency was set at 10kHz. (a) grid voltage of phase-A (V). 8 Simulation results of the dynamic response of HCC. and the hysteresis band width is set at 0. Lower: step change from 20A rms to 5A rms. the dc-link voltage was 365V. 9 Simulation results of the dynamic response of SVPWM-based PI control. Computer simulations have been conducted using PSIM simulation package for all three current control strategies. the switching dead time was 4uS. and with respect to the dynamic performance. For all the laboratory tests. the switching frequency is fixed at 10kHz. Lower: step change from 20A rms to 5A rms. 1314 . the switching frequency is limited up to 10kHz. the ADC sampling frequency was 40kHz. Upper: step change from 5A rms to 20A rms. 7 Simulation results of the steady-state response of current controllers with 20A rms current reference. the switching dead time is 4. In addition. the switching dead time was compensated in the SVPWM algorithms. Kp and Ki of PI regulators in SVPWM-based PI current controller are 4. Fig. the output filter inductance is 2mH per phase. while Figs. The simulation results showe that with distorted grid voltages. the dc-link voltage is 360V.
L. Comparison of output current THD for three PWM controllers Current Control Strategy Hysteresis Control SVPWM-Based PI Control SVPWM-Based Predictive Control 40 Current Waveform Fig.2% 1. It can be clearly seen that both advanced SVPWM-based current controllers yield much smoother output current waveforms than HCC. the mismatch mainly causes steady-state errors. the modeled inductance Lm was successively set B. Kp and Ki were chosen as 2 and 0. Table1. 13 to 15 present the experimental results of the dynamic responses for three current controllers respectively. (c) Phase-A output current controlled by SVPWM-based PI controller (A).75L. In the experiments. since it is built on the knowledge of the system model. It can be found that when Lm is less than L.01ms/digit (d) Fig. As well known. its performance is influenced by the mismatch of the system parameters. 200 at 0. It can be seen that with the well-designed error compensation and variable prediction in feedback and feed-forward loops. (b) Phase-A output current controlled by HCC (A). 12(d) 3 THD 3. 11 presents the results of steady state tests of the three current control strategies.01ms/digit. and the output currents with a reference of 15A (rms) were recorded for comparison. The THD of inverter output currents for the current controllers is given in Table 1.1% 1. 12 Experimental waveforms of the output currents (A) (left) and the corresponding current errors (A) (right) of SVPWM-based predictive controller with different modeled inductance.2% 100 0 current error in q-axis (A) 2 1 0 -1 -2 -3 -3 -100 20 -200 1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 0 (a) 40 20 -20 -40 1 401 801 1201 1601 -2 -1 0 1 2 3 0 current error in d-axis (A) 3 40 current error in q-axis (A) -20 2 1 0 -1 -2 -3 -3 3 20 -40 1 501 1001 1501 2001 (b) 2501 3001 3501 4001 4501 5001 0 40 20 -20 0 -40 1 401 801 1201 1601 -2 -1 0 1 2 3 current error in d-axis (A) -20 40 current error in q-axis (A) -40 1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 2 1 0 -1 -2 -3 -3 (c) 40 20 20 0 -20 0 -20 -40 1 1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 401 801 1201 1601 -2 -1 0 1 2 3 current error in d-axis (A) -40 Fig.75L. With effective current error compensation algorithms and predictors. Experimental tests on different modeled inductances have been conducted to investigate the influence of the filter inductance mismatch on the predictive controller. which are represented by the output filter inductance in this paper. (a) grid voltage of Phase-A (V). the mismatch causes the steady-sate error and also the current oscillation which degrades the system stability.02 respectively. Steady State Tests Figs. Lower: Upper: Lm=1. Upper: Lm=0. Time: 0.0%. Fig. and 1.A.25L. the inverter output currents under advanced SVPWM are highly sinusoidal even when the grid phase voltage has marked low-order harmonics with a voltage THD of 3. 12(c) Fig. where THD is measured up to the 51st harmonic. In the experiments. 1315 .25L. 12(b) Fig. the performance of PI controller depends on the proper choice of Kp and Ki. where current references are all 20A (rms) with a unity power factor. the output current reference of the inverter was step changed from 5A (rms) to 20A (rms) and then changed reversely after the inverter operated in steady state. Time: 0.11 Experimental results of the steady-state response of the three current controllers with 20A rms current reference. 12 shows the three-phase output currents with different Lm and their corresponding current errors in d-q reference frame. Middle: Lm=L. (d) Phase-A output current controlled by SVPWM-based predictive controller (A). For the SVPWM-based predictive controller. and when Lm is larger than L. Transient State Tests In order to evaluate the dynamic responses of three current controllers. the effect on system dynamic response due to the grid harmonics and the control delay have been greatly alleviated and that the dynamic responses of the two advanced SVPWM-based current control strategies are almost as good as that of HCC. Figs.
“Implementation and control of grid connected AC-DC-AC power converter for variable speed wind energy conversion system. Hahm. Kwon. “SVPWM-based current controller with grid harmonic compensation for three-phase grid-connected VSI. “Control and filter design of three-phase inverters for high power quality grid connection.” IEEE Thirty-First Annual Meeting of Industry Applications Conference 1996 (IAS '96). pp.” Proceeding of Eleventh Annual Applied Power Electronics Conference and Exposition 1996 (APEC '96). Lower: step change from 20A rms to 5A rms (A). Jan. A. March 1996. However. pp. Power Electronics. 752-760. Skudelny. Sudhoff. pp. G. “Development of adaptive hysteresis band current control strategy of PWM inverter with constant switching frequency. “Vector current controlled grid connected voltage source converter-influence of nonlinearities on the performance. pp. Chang. Industry Application. 2000. However. Song. 18.40. the controller is insensitive to system parameters since the algorithm does not need system models. pp. 2494 – 2500. 45. pp.Power Electron. pp. vol.. B. 40     20 0 -20 -40 1 40 501 1001 1501 2001 2501 3001 3501 4001 4501 5001  20 0   1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 -20 -40 Fig. L. Holmes. Vol.-K. Energy Information Administration. Martin. Upper: step change from 5A rms to 20A rms (A).  VIII..154-158. Jang-Hyoun Youm.” IEEE Trans. Krause. C. robustness.” IEEE Trans.-C. L. namely hysteresis control. E. pp. vol. 1. International Energy Outlook. J. S. April 2004. June 2002. Kojabadi. John Wiley & Sons. 2003.” IEEE PESC’04. vol.” IEEE PESC98. Twining and D. V. S. Electron. vol. D. low current ripple and highly sinusoidal waveform. 1316 . 14 Experimental results of the dynamic responses of SVPWM-based PI controller. Q. 1998. 1. Jan. van der Broeck.1. 906 – 913. M. Lower: step change from 20A rms to 5A rms (A). 691-703. “An improved space-vector-based hysteresis current controller. Choi. Chang and H. L.888-895.-D. Upper: step change from 5A rms to 20A rms (A).36 . Analysis of Electric Machinery. which limit its applications in grid-connected VSIs. Lower: step change from 20A rms to 5A rms (A). W. Upper: step change from 5A rms to 20A rms (A). “Review of interconnection standards for distributed power generation. May 2003. Vol. Green.01ms/digit.-W. CONCLUSION    Three current control strategies. Malesani. pp.” Large Engineering Systems Conference on Power Engineering 2002 (LESCOPE’ 02). 1998. D. Holmes. “Grid current regulation of a three-phase voltage source inverter with an LCL input filter. 2. G. B. Time: 0. vol. Song.40 20 0 -20 -40 1 40 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 20 0 -20 -40 1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 Fig. issue 1.” IEEE Canadian Conference on Electrical and Computer Engineering 1998 (CCECE’98). Moreover. REFERENCES  0 -20 -40 1 40 501 1001 1501 2001 2501 3001 3501 4001 4501 5001    20 0 -20 -40 1 501 1001 1501 2001 2501 3001 3501 4001 4501 5001 Fig. Power Electron. May 24-28 1998. TMS320LF/LC240x DSP Controllers Reference Guide: System and Peripheral. its performance is sensitive to system parameters. June 20-25 2004. SVPWM-based PI current control offers an excellent steady-state response featured by precise control with zero steady-state error. high current ripple and variable switching frequency are the drawbacks of HCC.” APEC’03. especially to the output filter inductance L which may vary due to its nonlinearity. 6-10 1996. 13 Experimental results of the dynamic responses of HCC. T. Oct. “Implementation of a direct digital predictive current controller for single and three phase voltage source inverters.-H. H. Svensson and M. “PWM control strategies for wind turbine inverters. Simulation and experimental results present the advantages and disadvantages of each current control strategies. Stanke. 1. P. October 1998. Chun. P. January-February 1988. O.531-537.” IEEE Trans. 45.309 – 312. Literature No: SPRU357. 2002.01ms/digit. 2nd edition. SVPWM-based PI control and SVPWM-based predictive control. pp. G.. Q. Kang. 194-199. 2003.M. Time: 0. Prodanovic and T. Ind. vol. M.” IEEE Trans. implemented and compared on a 30kW three-phase grid-connected PWM VSI by Authors. vol. extremely fast dynamic response and inherent over-current protection.4. C. vol. vol. 24. Its dynamic response is slightly inferior to those of HCC and SVPWM-based predictive current control even though this can hardly be distinguished in the experimental results.142-150. Chang. H. “Analysis and realization of a pulsewidth modulator based on voltage space vectors. 373-380. pp. and N. Oct. Kazmierkowski and L. Zeng. M. Wasynczuk and S. Texas Instruments. 15 Experimental results of the dynamic responses of SVPWM-based predictive controller. P. Min. SVPWM-based predictive current control also has a very good steady-state performance and at the same time. “Current control techniques for three-phase Voltage-Source PWM Converters: A survey.” IEEE Trans. have been developed. pp. 18. Lindgren.01ms/digit. D. provides a good dynamic performance.. Wang.. Time: 0. 40 20 HCC is attractive for its highly simplicity.
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 listening from where you left off, or restart the preview.