IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO.

3, MARCH 2012

1443

Generalized Design of High Performance Shunt Active Power Filter With Output LCL Filter
Yi Tang, Student Member, IEEE, Poh Chiang Loh, Member, IEEE, Peng Wang, Member, IEEE, Fook Hoong Choo, Feng Gao, Member, IEEE, and Frede Blaabjerg, Fellow, IEEE

AbstractThis paper concentrates on the design, control, and implementation of an LCL-lter-based shunt active power lter (SAPF), which can effectively compensate for harmonic currents produced by nonlinear loads in a three-phase three-wire power system. With an LCL lter added at its output, the proposed SAPF offers superior switching harmonic suppression using much reduced passive ltering elements. Its output currents thus have high slew rate for tracking the targeted reference closely. Smaller inductance of the LCL lter also means smaller harmonic voltage drop across the passive output lter, which in turn minimizes the possibility of overmodulation, particularly for cases where high modulation index is desired. These advantages, together with overall system stability, are guaranteed only through proper consideration of critical design and control issues, like the selection of LCL parameters, interactions between resonance damping and harmonic compensation, bandwidth design of the closed-loop system, and active damping implementation with fewer current sensors. These described design concerns, together with their generalized design procedure, are applied to an analytical example, and eventually veried by both simulation and experimental results. Index TermsActive power lter, current control, LCL lter, resonance damping.

I. I NTRODUCTION HE higher order LCL lter has commonly been used in place of the conventional L-lter to give a better smoothing of output currents from a voltage source converter [1], [2]. Its applications to grid-connected inverters and pulsewidth modulated active rectiers have recently attracted a lot of research attentions [1][8], mainly due to its ability to minimize the amount of current distortion injected into the utility grid. Power quality of the grid is hence enhanced, which is particularly important for small-scale distributed generation systems, where the ac bus is not strong [6]. Despite having these advantages, there are some challenges faced by the LCL lter in practical implementations, whose common concern is the

Manuscript received March 29, 2011; revised July 12, 2011; accepted August 16, 2011. Date of publication September 6, 2011; date of current version October 25, 2011. Y. Tang, P. C. Loh, P. Wang, and F. H. Choo are with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 (e-mail: tang0175@ntu.edu.sg; pcloh@ieee.org; epwang@ntu.edu.sg; efhchoo@ntu.edu.sg). F. Gao is with the School of Electrical Engineering, Shandong University, Jinan 250061, China (e-mail: fgao@sdu.edu.cn). F. Blaabjerg is with the Institute of Energy Technology, Aalborg University, 9220 Aalborg East, Denmark (e-mail: fbl@iet.aau.dk). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TIE.2011.2167117

possibility of exciting serious resonance at certain frequency. The overall system might therefore be unstable, but fortunately can be resolved by applying existing damping techniques, like adding a real resistor in series with the lter capacitor [1], actively feeding back some measured or estimated electrical variables for control purposes [7][12], and splitting the lter capacitor to reduce the system order [13], [14]. These techniques have no doubt contributed prominently to the widespread adoption of LCL lter by the industry [1]. Although extensive, most investigations on LCL lter now have focused on topics like fundamental current tracking and resonance damping for mostly simple grid-tied dc-ac inverters and ac-dc rectiers. For these applications, the LCL resonance frequency is usually tuned to be at least ten times of the fundamental line frequency [1]. It is therefore not difcult to simultaneously achieve the desired control objectives at fundamental frequency using existing control techniques. In [15], a slightly more complicated design scenario was considered for a three-phase inverter driven by direct power control and connected to an ac grid with fth harmonic voltages. A similar study was conducted in [16], where the low-order grid harmonic problem was solved by feeding forward the grid voltages. The highest harmonic orders considered in these studies were, however, quite small and hence far away from the resonance frequencies of their respectively designed LCL lters. Their effectiveness in compensating harmonics was, therefore, quite expected since interactions with system damping and stability were not signicant. Other more challenging studies on LCL lter can certainly be found in the literature, but only a few has discussed about merits and design challenges faced when applied to shunt active power lter (SAPF) [17], [18]. In [19], a repetitive control scheme coupled to a one-beat-delay current controller was proposed for LCL-lter-based SAPF. Although the system exhibited good performance under both dynamic and steady-state conditions, the design of LCL lter and resonance damping control was not specically addressed. A design procedure for determining the parameters of the LCL lter was subsequently discussed in [20] for SAPF controlled by a hysteresis scheme with variable switching frequency. The provided methodology, however, led to very different grid- and converter-side inductors, which were therefore not yet optimized, based on concepts discussed in [21], [22]. Moreover, only passive damping technique was considered in [20], meaning real resistors were added in series with the lter capacitors, leading to unnecessary power losses. These earlier studies on SAPF also included only simulation results with their practicalities left unveried.

0278-0046/$26.00 2011 IEEE

1444

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 3, MARCH 2012

Fig. 1. Circuit diagram of proposed SAPF with output LCL lter.

A comparative study can also be found in [23], where the LCL-lter-based and L-lter-based SAPFs were compared experimentally. Conceptual explanation for that study was, however, lacking, hence making it hard to appreciate improvement contributed by the LCL lter. These shortfalls, together with a number of other issues identied in the literature, lead to the general belief that LCL lter for the more complex SAPF has not been fully understood. Its real advantages over the L-lter-based SAPF are, therefore, not yet well-dened and are investigated here by addressing a few control objectives related to the LCL-lter-based SAPF. As a start, a frequency domain model of LCL lters, which takes into account the phase lag introduced by the LCL resonance, is established. Analysis of this model reveals that the maximum achievable system bandwidth is closely linked to the resonance frequency of the LCL lter. A general design guideline is then proposed to ensure proper placement of the resonance frequency within an appropriate chosen range, so as to simultaneously achieve accurate harmonic compensation and optimum resonance damping without any tradeoff noted between them. Active damping, being more efcient, is also explicitly considered here for embedding within the control loop, so as to alter the plant transfer function to get a more welldamped system. A design example then follows, whose objective is to compensate for harmonic currents up to the 25th order in a 50-Hz three-phase three-wire power system.

The designed system is next compared with its L-lter-based correspondence, designed to have similar ripple ltering. Through the comparison, an attractive advantage of the LCLlter-based SAPF is identied for cases where high modulation index (e.g., 0.9) is demanded, which so far has not been previously discussed. Simulation and experimental results for verifying this advantage, together with other performance improvements, are subsequently provided, before concluding on the effectiveness of the proposed SAPF design and control methodology. II. M ODEL OF LCL F ILTER Fig. 1 shows the typical circuit diagram of a SAPF implemented as a three-phase three-wire system. Between the SAPF and utility grid is an LCL lter added for current smoothing, whose model is formulated by rst making a few assumptions for simplifying the analysis. Foremost would be to assume that the three-phase voltages at the point of common coupling are sinusoidal and balanced. That then means the grid can reasonably be treated as a short-circuit when performing stability analysis in the high frequency range. In addition, all equivalent series resistances (ESRs) of passive components, including the converter-side inductor Lf f , grid-side inductor Lgf , and lter capacitor Cf , are neglected, since they provide some degrees of resonance damping, and would thus raise the overall system stability. Ignoring ESRs therefore represents the worst case

TANG et al.: GENERALIZED DESIGN OF HIGH PERFORMANCE SHUNT ACTIVE POWER FILTER WITH OUTPUT LCL FILTER

1445

Fig. 3. Bode plot of (6) obtained using the parameters listed in Table I. Fig. 2. Plant models of (a) undamped and (b) actively damped SAPFs. TABLE I S YSTEM PARAMETERS U SED FOR S IMULATION AND E XPERIMENT

scenario in terms of damping, even though it represents the best case in terms of loss reduction. Applying these assumptions then leads to the model shown in Fig. 2(a) for representing the power stage of the SAPF, whose converter bridge has been represented by a gain of Vdc /2, as per previous practice. Fig. 2(a) can further be written as a set of transfer functions in the frequency domain, as demonstrated by Gp (s) = Vdc /2 Igf (s) = Vm (s) Lf f Lgf Cf s3 + (Lf f + Lgf )s (1) (2)

Icf (s) = Lgf Cf s2 Igf (s)

where Vm is the normalized modulating signal, Vdc is the dc-link voltage, Igf and Icf are the currents injected into the grid and absorbed by the lter capacitor, respectively. Applying unity feedback control to (1) then results in a characteristic polynomial without the s2 term. The overall closed-loop system is thus unstable or marginally stable according to the Rouths stability criterion and would need additional damping to stabilize it. Passive damping is certainly a simple and straightforward method for consideration but would introduce high losses, particularly for SAPF, where high switching harmonic current will ow through the added damping resistor per phase. Passive damping is therefore not considered further. Instead, active damping is focused, which when applied, leads to the actively damped system shown in Fig. 2(b). In that gure, the lter capacitor current is sensed and added to the modulating signal through a damping gain Kd . The new transfer function is then written as Igf (s) Vdc /2 . (3) = Vm (s) Icf (s)Kd Lf f Lgf Cf s3 + (Lf f + Lgf )s Upon substituting (2) for Icf (s), the new plant model Gpd (s) is derived as Gpd (s) Igf (s) = Vm (s) = Vdc /2 . (4) Lf f Lgf Cf s3 +(Vdc /2)Kd Lgf Cf s2 +(Lf f +Lgf )s

With a nite s2 term now introduced to its denominator, closed-loop stability of (4) can easily be tuned by varying the feedback gain Kd appropriately. Damping factor , representing the extent of resonance damping, can also be tuned as 2res = Vdc Kd = 2 2Lf f Lf f + Lgf Lf f Lgf Cf (5)

where res represents the LCL resonance frequency, determined solely by the passive component values. Substituting (5) into (4) then results in the following simplied damped system model: Gpd (s) = Vdc 1 . 2 2Lf f Lgf Cf s (s2 + 2res s + res ) (6)

Bode diagrams of (6) under various damping factors can now be plotted as in Fig. 3, using parameters listed in Table I. Also drawn are the rst-order curves associated with a simple L-lter, whose inductance is set to Lt = Lf f + Lgf .

1446

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 3, MARCH 2012

Comparing the curves then leads to the general observation that the magnitude response of the LCL lter approaches that of the L-lter in the low frequency range (below resonance frequency), regardless of the damping factor chosen. Their phase responses can, however, be quite different, which may signicantly impact the system stability. Moreover, it is sensed that the control bandwidth cannot exceed or even be close to the resonance frequency, since it will create a 180 phase lag, whose resulting inuence is an insufcient phase margin for closed-loop control. Non-negligible delays caused by sampling and modulation would also constrain the system crossover frequency c , limiting it to be less than 0.3 times the resonance frequency res , based on the region suggested in Fig. 3. Dening as c /res 0.3 then leads to a phase lag of LCL radians for the LCL lter at c , calculated as LCL ={Gpd (jc )} = = Vdc 1 3 +2 2 2 2Lf f Lgf Cf (jc ) res (jc ) +res jc Vdc 2Lf f Lgf Cf 1 2 (jres )3 +2res (jres )2 +res jres
Fig. 4. Root locus of (6) showing trajectories of closed-loop poles.

gain is, therefore, absent, implying that optimal damping of the LCL lter cannot be guaranteed. This is clearly conrmed by the dynamic results presented in [25], where obvious current oscillations and overshoots can be seen. Proper sizing of the damping gain is, therefore, important and is precisely determined here by using the plant model in (4) and (5) to arrive at Kd = 4 Vdc Lf f (Lf f + Lgf ) Lgf Cf (8)

Vdc 1 = 3 3 2Lf f Lgf Cf 22 res +(3 )jres 12 . =arctan 2 (7)

When the damping factor of LCL resonance is chosen to be 0.707 (a very typical value used for second-order system), LCL is calculated as 0.64 or 115 , which is not a very severe degradation, as compared to the value of 0.5 or 90 of an L-lter. Parameter c or 0.3res should, moreover, be chosen much higher than the highest harmonic current compensated by the SAPF, so as to completely eliminate its interaction with the LCL resonance. The desired switching frequency can eventually be set, which according to [24], should be set at least two times of res , so as to provide sufcient harmonic attenuation around the switching frequency. III. C ONTROL OF SAPF Upon guring out the relationships among system bandwidth, LCL resonance frequency, and converter switching frequency, development of control algorithms for the LCLlter-based SAPF can proceed. The resulting control block diagram is shown in Fig. 1, where active damping, harmonic current compensation, and dc-link voltage control are all considered. Regarding active damping, it has earlier been demonstrated on LCL-lter-based SAPF in [25], where the lter capacitor current was measured and fed to the inner current feedback control loop to imitate a virtual damping resistor. Although the controlled system can eventually be stabilized in the steady state, its damping gain at resonance frequency was not analytically determined, but rather obtained through trial and error. Proper guide on how to choose the damping

where Kd is shown to be proportional to the damping factor , whose value can freely be tuned to arrive at the desired damping response. Referring next to (6), the general observation noted is that the damped LCL network has a pair of conjugate complex poles in the left half plane and another pole at the origin. The locations of these poles, and hence the system response and performance, can be varied by applying unity feedback, and then tuning the proportional gain Kpc of the current controller. Theoretically, a large proportional gain would force the real closed-loop pole to move further away from the imaginary axis, hence attenuating its impact, and causing the two conjugate poles to dominate. The conjugate poles would also move closer toward the right half plane, causing the system to gradually become unstable. The described pole trajectories are shown by the root locus of (6) plotted in Fig. 4, which also shows the closeness of the poles, even though they are gradually moving apart, as Kpc increases. The impact of the real pole, therefore, cannot be neglected, but should be understood by comparing the step response of the closed-loop system with the following secondorder system Gs (s) at different damping factors: Gs (s) =
2 res . 2 s2 + 2res s + res

(9)

The results are plotted in Fig. 5, where a feature noted with the real pole is its slowing down of the overall step response, hence leading to a longer settling time. It, however, damps transient oscillations more forcefully, leading to smaller overshoots at all damping factors. Based on these observations, a recommended value for with active damping incorporated is 0.7, since it leads to a smooth system recovery with no

TANG et al.: GENERALIZED DESIGN OF HIGH PERFORMANCE SHUNT ACTIVE POWER FILTER WITH OUTPUT LCL FILTER

1447

however, close to each other, allowing their effects to cancel out. The overall system response would, therefore, still be dominated by those three poles indicated in Fig. 4. Adding on to the fundamental current regulation is the compensation for harmonic currents, which can be implemented in the synchronous dq frame by using multiple paralleled resonant controllers, expressed as Gch (s) =
Fig. 5. Step responses of Gs (s) and closed-loop of Gpd (s).

Kih s s2 + (6nn )2 n=1

(13)

noticeable overshoot and an acceptably short response time. At times, can also be chosen slightly smaller, like 0.5, to introduce a safer phase margin (see Fig. 3), but slightly higher overshoot. Further reduction of is, however, not recommended, since it will not improve the phase response too much, but only lead to even serious oscillations, as seen in Fig. 5. Upon deciding on the value of , Kd can follow up be determined using (8) without much difculty. Proceeding on to the inner current control loop, its responsibilities are claried as power ow regulation and harmonic current compensation. Unlike those methods proposed in [19], [23], [25], where both nonlinear load and SAPF currents were sensed, only the line currents are measured here, and controlled to be balanced and sinusoidal. One set of current sensors, together with the load harmonic extraction module for generating the SAPF reference currents, are therefore eliminated, leading to a simpler control implementation. For unity power factor operation, the reactive grid current reference should further be set to zero, while using the following proportional-integral (PI) controller in the synchronous dq frame for forcing the grid current to track an active reference per phase: Gcf (s) = Kpc 1 + 1 c s (10)

where Kpc and c are, respectively, the proportional gain and integral time constant. As Kpc dominates the bandwidth of the inner current control loop, its value must properly be chosen to ensure that c falls well within the frequency range discussed in Section II, and indicated in Fig. 3. For that, the following equation can be used for calculating Kpc [26], after deciding on the value of c (= 0.3res , as explained earlier): Kpc c (Lf f + Lgf ) c Lt = . Vdc /2 Vdc /2 (11)

where n , Kih , and k represent the nominal angular frequency, respective resonant gain, and highest harmonic order that can be compensated. Rather than (13), resonant controllers placed in the stationary frame can also be used, but would result in more terms for summation. This is explained in [27], where it is shown that a resonant controller in the synchronous frame is more effective, since it represents two equivalent resonant terms in the stationary frame for compensating two harmonics. Other details on the discretization and optimization of the resonant controllers can be found in [27], [28], and are therefore not explicitly elaborated here. Returning back to k in (13), its value should rightfully be much smaller than the specied resonance frequency. That then corresponds to the low frequency range shown in Fig. 3, whereas mentioned earlier, would cause the LCL lter to behave like a small inductance. Output current produced by the SAPF in this range can, therefore, be of high slew rate for tracking the harmonic reference accurately. A rst-order L-lter, designed to produce the same switching ripple ltering, would not be able to respond that fast because of its larger inductance. A possible solution for it is to tune up its controller proportional gain to maximize the system bandwidth. This, however, is quite hard to achieve in reality, since it can lead to unwanted noise amplication, and hence instability. Moving next to the outer voltage control loop of the SAPF, its responsibility is to keep the dc-link capacitor voltage constant by compensating for active power losses in the system. Its realization can, therefore, be just the stationary-frame PI controller Gv (s) written in (14), whose high dc gain would force the tracking error to zero Gv (s) = Kpv 1 + 1 v s (14)

The value of c , on the other hand, mainly determines the fundamental steady-state tracking error and should therefore be small. A conservative value recommended for it to avoid impacting on system stability, like the reduction of phase margin, is given in {Gcf (jc )} = 90 arctan c c 2 and c 30/c . (12)

It should also be noted that adding this PI controller in the current control loop will introduce an additional pair of zero and pole to the closed-loop response. These added terms are,

where Kpv is the proportional gain, and v is the time constant of the integral term. Gv (s) here must rightfully be designed to be much slower than the current controller Gc (s)(= Gcf (s) + Gch (s)), so as to avoid interference between them. Another feature exhibited by the dc-link control loop is its lower dc voltage requirement. This is possible for the LCL lter, since its total inductance is much smaller, and hence the unltered converter output needs to compensate for only a much smaller voltage drop across it. The dc-link voltage can, therefore, be reduced, while yet maintaining the same high modulation index that is less prone to overmodulation. This is, however, not true for the L-lter-based SAPF, which always requires a larger inductance, and hence a larger voltage drop across it. Consequently, its output current would run

1448

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 3, MARCH 2012

Fig. 6. Simulated results of LCL-lter-based SAPF when subjected to a 33% to 100% step-up load change.

Fig. 7. Simulated results of L-lter-based SAPF when subjected to a 33% to 100% step-up load change.

into saturation way earlier than that of an LCL-lter-based SAPF system. Being applied to SAPF also leads to a more prominent voltage drop, since it deals with sizable harmonic currents owing through sizable inductive impedance that is k times larger than the fundamental. In contrast, similar effects experienced by grid-tied inverters and rectiers would likely be less obvious, since they deal with fundamental currents and inductive impedances only. IV. G ENERALIZED D ESIGN P ROCEDURE AND E XAMPLE Based on reasons discussed earlier about parameter sizing, the generalized design procedure for LCL-lter-based SAPF is now presented in per-unit terms for easier future system scalability. The procedure would quite expectedly begin by dening the highest order harmonic current k that needs to be compensated. According to Section II, the cutoff frequency c and resonance frequency res should then be larger than kn and kn /0.3, respectively (n being earlier dened as the nominal fundamental frequency). To be slightly more conservative so as to fully avoid interference between harmonic current compensation and resonance damping, the resonance frequency can instead be set to res = kn /0.25 = 4kn . (15)

Fig. 8.

Experimental steady-state results of LCL-lter-based SAPF.

where Zb is the base impedance calculated by dividing the system rated power from the square of its rated voltage. Noting also that the grid and converter-side inductances should be set equal to produce the lowest resonance frequency, and hence the maximum switching harmonic attenuation, the set of lter parameters recommended should be Lf f = Lgf = xLb = xZb /n Cf = yCb = y/n Zb (17)

To next cater for sufcient switching harmonic suppression, the desired switching frequency of the converter can be set two times larger than the LCL resonance frequency. Even higher switching frequency can of course be used for better harmonic attenuation, but would incur additional losses, and therefore not rmly recommended. With res now decided, the lter component values can follow up be determined, but before progressing ahead, some base inductance Lb and capacitance Cb values need to be dened for per-unit representation. These bases are written as Lb = Zb /n Cb = 1/n Zb (16)

where x and y are the per unit inductance and capacitance, respectively, which are usually chosen below 5% to bring down the system cost. Substituting (17) to (5), followed by some simple manipulation, then leads to xy = 2(n /res )2 = 2(1/4k)2 . (18)

Apparently, (18) has innite solutions, but it is recommended here that the total per unit value of the two inductors should be chosen equal to that of the lter capacitor, hence yielding Lf f = Lgf = (1/4k)Lb Cf = (1/2k)Cb . (19)

TANG et al.: GENERALIZED DESIGN OF HIGH PERFORMANCE SHUNT ACTIVE POWER FILTER WITH OUTPUT LCL FILTER

1449

Fig. 9.

Experimental results of LCL-lter-based SAPF when subjected to a 33% to 100% step-up load change.

The closest commercially available capacitance value should then be chosen, while the inductance value can be wound accordingly. If desired, the actual realized resonance frequency can be calculated using these nal chosen reactive values and should roughly match the value recommended in (15) for the design. The reactive values, together with the chosen damping factor (0.5 0.7), cutoff frequency c , and dc-link voltage Vdc , can next be substituted to (8), (11), and (12) to determine the important control parameters of Kd , Kpc , and c , respectively. Other control parameters like Kih in (13) for harmonic current compensation, and Kpv and v in (14) for dclink voltage regulation, can be tuned based on classical outer loop control and zero tracking error operating principles. These are well-established and, therefore, not explicitly shown here. The above design procedure can now be applied to an example SAPF, whose objective is to compensate up to the 25th harmonic current (k = 25) in a 173-V (line-to-line RMS), 2-kVA, 50-Hz three-phase three-wire system. From (15), the resonance frequency should rightfully be chosen as 1250/0.25 = 5000 Hz or a value between 1250/0.3 = 4167 Hz and 5000 Hz. Since the desired switching frequency should be at least two times larger than the resonance frequency, its value should be xed to 10 kHz or more. Using (16), the base inductance and capacitance values can next be calculated as 47.8 mH and 212 F, respectively. Substituting these bases and k = 25 to (19) then leads to the calculated values of 0.478 mH and 4.24 F for the LCL lter. Based on available commercial values, the nal LCL parameters are chosen as Cf = 5 F and Lf f = Lgf = 0.5 mH, giving a recalculated resonance frequency of fres = res /2 = 4504 Hz, which indeed falls within the range from 4167 Hz to 5000 Hz that has earlier been identied. Using these reactive parameters, together with = 0.5, c = 8485 rad/s, and Vdc = 300 V, the corresponding control parameters are eventually determined as Kd = 0.09, Kpc = 0.05, c = 0.0035 s, Kih = 20, Kpv = 0.2, and v = 0.04 s. V. S IMULATION R ESULTS Simulation was conducted with Matlab/Simulink and PLECS, based on the system shown in Fig. 1. The aim was to examine the performance of the proposed SAPF system

Fig. 10. Grid current spectrum versus IEEE 519-1992 standard.

Fig. 11. Experimental steady-state results of L-lter-based SAPF.

and to compare it with the conventional L-lter-based SAPF. The specications of the designed SAPF and other system parameters were based on those derived in Section IV and summarized in Table I for comprehensiveness. The ac line and dc-link voltages stated in Table I were found to give an approximate modulation index of 141/150 = 0.94 for the SAPF converter, which indeed was a reasonably high value. Such a high value was used for proving the lower likelihood of entering overmodulation by the LCL-lter-based SAPF. Beginning with Fig. 6 where the proposed SAPF was subjected to a 33% to 100% step-up nonlinear load change, the results show a smooth compensation of harmonic currents up to the 25th order. Total harmonic distortion (THD, calculated

1450

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 3, MARCH 2012

Fig. 12. Experimental results of L-lter-based SAPF when subjected to a 33% to 100% step-up load change.

Fig. 13. Harmonic spectrum of grid current in high frequency range.

up to the 100th order) of the full load grid current is in fact only 2.91%, even though the load THD is 44.20%. Modulating references in the last plot of Fig. 6 are also noted to vary below unity without entering overmodulation, even with a detected dip in dc-link voltage during the transient. For comparison, the LCL network was next replaced by a 5 mH inductor, followed by a retuning of the current controller gains to produce the same dynamic response, as per the previous simulation. Subjecting the L-lter-based SAPF to the same load transient then results in Fig. 7, where a prominent increase in modulation index is observed with larger power ow. The increase is for compensating the larger voltage drop across the L-lter and transient dip in dc-link voltage. That pushes the system into overmodulation like the example illustrated here, whose eventual effect is a set of distorted grid currents with THD as high as 7.79%. VI. E XPERIMENTAL R ESULT To validate the proposed SAPF practically, a prototype was built in the laboratory, based on the same parameter values listed in Table I and circuit connection shown in Fig. 1, as per earlier used for simulation. The developed control algorithm was executed on a dSPACE1103 real-time platform, and the digital signal processor TMS240F2812 was employed with its sampling frequency being set as 20 kHz in the controller board, which is twice of the switching frequency. The rst set of experimental results obtained is shown in Fig. 8, during which the proposed SAPF is supplying 100% nominal load power

in the steady state. As anticipated, the SAPF compensates the load harmonic currents well, giving rise to smooth sinusoidal currents drawn from the grid with a THD of only 3.49%. This is despite the load currents being heavily distorted with a THD of 40.37%. Also shown in the last plot of Fig. 8 is the modulating reference for one phase of the SAPF, being fully within the shaded linear modulation range. The same demanded performance is shown in Fig. 9, where a step load transient from 33.3% to 100% of nominal load is intentionally triggered. This sudden increase in harmonic currents causes the dc-link voltage Vdc to drop to 279.2 V, which, due to limited channels available on the digital scope, is not explicitly shown here. This drop under an undisturbed ac grid would cause the modulation index to rise to approximately 141/139.6 1.01, which, although is slightly greater unity, would still not lead to overmodulation problem. The reason is partly due to the low series impedance of the LCL lter and partly due to the presence of triplen offset that can clearly be seen added to the modulating reference of each phase. Transient events are therefore mostly ridden through smoothly with no obvious harmonic distortion sensed from the line current. For further demonstrating that the proposed SAPF meets international standard, low-order harmonic spectrum of the resulting grid current under nominal load condition is plotted in Fig. 10, which undeniably shows that the IEEE 519-1992 standard has been met comfortably. The output lter of the SAPF was subsequently replaced by a 5-mH inductor, and retested under the same experimental conditions. Fig. 11 shows the steady-state experimental results, where slight overmodulation is observed under nominal load operating condition. This overmodulation is caused by the need to compensate for a larger voltage drop across the lter impedance, while yet keeping the same low dc-link voltage. Its extent can in fact worsen under transient conditions, like the 33% to 100% load step change shown in Fig. 12. The corresponding effect is of course a distorted grid current, whose low-order spectrum under nominal load condition is shown in Fig. 10. The spectrum undoubtedly shows poorer performance than the IEEE standard with its THD of 7.86% in excess of the normal 5% limit. Adding on to these inferiorities is its poorer switching

TANG et al.: GENERALIZED DESIGN OF HIGH PERFORMANCE SHUNT ACTIVE POWER FILTER WITH OUTPUT LCL FILTER

1451

ripple ltering, even after increasing its lter inductance to be ve times larger than that of the LCL lter. The corresponding results are shown in Fig. 13, where the switching and associated sideband harmonics are plotted for both L-ltered and LCLltered SAPFs. The results indeed show the former having poorer ltering performance, particularly at high frequency. VII. C ONCLUSION This paper proposes an LCL-lter-based SAPF for threephase three-wire power system, together with its generalized design and control procedure. Being of higher order, the proposed SAPF provides better ltering without using large passive components. Its resulting output currents therefore have high slew rate for accurate harmonic compensation without unnecessarily entering overmodulation mode during transient. Stability concerns are, however, more involved but can be resolved by the proposed design and control tuning methodology. Experimental testing of the methodology has proven its effectiveness in ensuring proper damping, overall stability, and smooth transient response achieved through a more comprehensive and generalized design procedure. R EFERENCES
[1] M. Liserre, F. Blaabjerg, and S. Hansen, Design and control of an LCLlter-based three-phase active rectier, IEEE Trans. Ind. Appl., vol. 41, no. 5, pp. 12811291, Sep./Oct. 2005. [2] A. A. Rockhill, M. Liserre, R. Teodorescu, and P. Rodriguez, Grid-lter design for a multimegawatt medium-voltage voltage-source inverter, IEEE Trans. Ind. Electron., vol. 58, no. 4, pp. 12051217, Apr. 2011. [3] E. Figueres, G. Garcera, J. Sandia, F. Gonzalez-Espein, and J. C. Rubio, Sensitivity study of the dynamics of three-phase photovoltaic inverters with an LCL grid lter, IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 706717, Mar. 2009. [4] P. Channegowda and V. John, Filter optimization for grid interactive voltage source inverters, IEEE Trans. Ind. Electron., vol. 57, no. 12, pp. 41064114, Dec. 2010. [5] R. Turner, S. Walton, and R. Duke, Stability and bandwidth implications of digitally controlled grid-connected parallel inverters, IEEE Trans. Ind. Electron., vol. 57, no. 11, pp. 36853694, Nov. 2010. [6] I. J. Gabe, V. F. Montagner, and H. Pinheiro, Design and implementation of a robust current controller for VSI connected to the grid through an LCL lter, IEEE Trans. Power Electron., vol. 24, no. 6, pp. 14441452, Jun. 2009. [7] E. Twining and D. G. Holmes, Grid current regulation of a three-phase voltage source inverter with an LCL input lter, IEEE Trans. Power Electron., vol. 18, no. 3, pp. 888895, May 2003. [8] F. Liu, Y. Zhou, S. Duan, J. Yin, B. Liu, and F. Liu, Parameter design of a two-current-loop controller used in a grid-connected inverter system with LCL lter, IEEE Trans. Ind. Electron., vol. 56, no. 11, pp. 44834491, Nov. 2009. [9] E. Wu and P. W. Lehn, Digital current control of a voltage source converter with active damping of LCL resonance, IEEE Trans. Power Electron., vol. 21, no. 5, pp. 13641373, Sep. 2006. [10] M. H. Bierhoff and F. W. Fuchs, Active damping for three-phase PWM rectiers with high-order line-side lters, IEEE Trans. Ind. Electron., vol. 56, no. 2, pp. 371379, Feb. 2009. [11] M. Malinowski and S. Bernet, A simple voltage sensorless active damping scheme for three-phase PWM converters with an LCL lter, IEEE Trans. Ind. Electron., vol. 55, no. 4, pp. 18761880, Apr. 2008. [12] P. C. Loh and D. G. Holmes, Analysis of multiloop control strategies for LC/CL/LCL-ltered voltage-source and current-source inverters, IEEE Trans. Ind. Appl., vol. 41, no. 2, pp. 644654, Mar./Apr. 2005. [13] G. Shen, D. Xu, L. Cao, and X. Zhu, An improved control strategy for grid-connected voltage source inverters with an LCL lter, IEEE Trans. Power Electron., vol. 23, no. 4, pp. 18991906, Jul. 2008. [14] G. Shen, X. Zhu, J. Zhang, and D. Xu, A new feedback method for PR current control of LCL-lter-based grid-connected inverter, IEEE Trans. Ind. Electron., vol. 57, no. 6, pp. 20332041, Jun. 2010.

[15] L. A. Serpa, S. Ponnaluri, P. M. Barbosa, and J. W. Kolar, A modied direct power control strategy allowing the connection of three-phase inverters to the grid through LCL lters, IEEE Trans. Ind. Appl., vol. 43, no. 5, pp. 13881400, Sep./Oct. 2007. [16] X. Wang, X. Ruan, S. Liu, and C. K. Tse, Full feedforward of grid voltage for grid-connected inverter with LCL lter to suppress currrent distortion due to grid voltage harmonics, IEEE Trans. Power Electron., vol. 25, no. 12, pp. 31193127, Dec. 2010. [17] H. Akagi, Y. Kanazawa, and A. Nabae, Instantaneous reactive power compensators comprising switching devices without energy storage components, IEEE Trans. Ind. Appl., vol. IA-20, no. 3, pp. 625630, May 1984. [18] H. Akagi, New trends in active power lters for power conditioning, IEEE Trans. Ind. Appl., vol. 32, no. 6, pp. 13121322, Nov./Dec. 1996. [19] Z. Qiu, J. Kong, and G. Chen, A novel control approach for LCL-based shunt active power lter with high dynamic and steadystate performance, in Proc. IEEE Power Electron. Spec. Conf., 2008, pp. 33063310. [20] M. T. Bina and E. Pashajavid, An efcient procedure to design passive LCL-lters for active power lters, Elect. Power Syst. Res., vol. 79, no. 4, pp. 606614, Apr. 2009. [21] K. Jalili and S. Bernet, Design of LCL lters of active-front-end twolevel voltage source converters, IEEE Trans. Ind. Electron., vol. 56, no. 5, pp. 16741689, May 2009. [22] Y. Tang, P. C. Loh, P. Wang, F. H. Choo, and K. K. Tan, Improved onecycle-control scheme for three-phase active rectiers with input inductorcapacitor-inductor lters, IET Power Electron., vol. 4, no. 5, pp. 603 614, May 2011. [23] O. Vodyakho and C. C. Mi, Three-level inverter-based shunt active power lter in three-phase three-wire and four-wire Systems, IEEE Trans. Power Electron., vol. 24, no. 5, pp. 13501363, May 2009. [24] J. Dannehl, C. Wessels, and F. W. Fuchs, Limitations of voltage-oriented PI current control of grid-connected PWM rectiers with LCL lters, IEEE Trans. Ind. Electron., vol. 56, no. 2, pp. 380388, Feb. 2009. [25] M. Routimo and H. Tuusa, LCL type supply lter for active power lterComparison of an active and a passive method for resonance damping, in Proc. IEEE Power Electron. Spec. Conf., 2007, pp. 29392945. [26] D. G. Holmes, T. A. Lipo, B. P. McGrath, and W. Y. Kong, Optimized design of stationary frame three-phase ac current regulators, IEEE Trans. Power Electron., vol. 24, no. 11, pp. 24172426, Nov. 2009. [27] R. Teodorescu, F. Blaabjerg, M. Liserre, and P. C. Loh, Proportionalresonant controllers and lters for grid-connected voltage-source converters, Proc. Inst. Elect. Eng.Elect. Power Appl., vol. 153, no. 5, pp. 750 762, Sep. 2006. [28] A. G. Yepes, F. D. Freijedo, J. Doval-Gandoy, O. Lopez, J. Malvar, and P. Fernandez-Comesana, Effects of discretization methods on the performance of resonant controllers, IEEE Trans. Power Electron., vol. 25, no. 7, pp. 16921712, Jul. 2010.

Yi Tang (S10) received the B.Eng. degree in electrical engineering from Wuhan University, Wuhan, China, in 2007 and the M.Sc. degree from Nanyang Technological University, Singapore, in 2009, where he is currently working toward the Ph.D. degree in the School of Electrical and Electronic Engineering. During the summer of 2007, he was a visiting scholar with the Institute of Energy Technology, Aalborg University, Aalborg East, Denmark, where he worked on the control of grid-interfaced inverters and uninterruptible power supplies.

1452

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 3, MARCH 2012

Poh Chiang Loh (S01M04) received the B.Eng. (Hons.) and M.Eng. degrees in electrical engineering from the National University of Singapore, Singapore, in 1998 and 2000, respectively, and the Ph.D. degree in electrical engineering from Monash University, Clayton, Vic., Australia, in 2002. During the summer of 2001, he was a visiting scholar with the Wisconsin Electric Machine and Power Electronics Consortium, University of Wisconsin-Madison, Madison, where he worked on the synchronized implementation of cascaded multilevel inverters and reduced common mode carrier-based and hysteresis control strategies for multilevel inverters. From 2002 to 2003, he was a project engineer with the Defence Science and Technology Agency, Singapore, managing major defense infrastructure projects and exploring new technology for defense applications. From 2003 to 2009, he was an assistant professor with the Nanyang Technological University, Singapore, and since 2009, he is an associate professor at the same university. In 2005, he has been a visiting staff rst at the University of Hong Kong, Hong Kong, and then at Aalborg University, Aalborg East, Denmark. In 2007 and 2009, he again returned to Aalborg University, rst as a visiting staff working on matrix converters and the control of grid-interfaced inverters, and then as a guest member of the Vestas Power Program.

Feng Gao (S07M09) received the B.Eng. and M.Eng. degrees in electrical engineering from Shandong University, Jinan, China, in 2002 and 2005, respectively, and the Ph.D. degree from the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, in 2009. From 2008 to 2009, he was a Research Fellow in Nanyang Technological University. Since 2010, he joined the School of Electrical Engineering, Shandong University, where he is currently a Professor. From September 2006 to February 2007, he was a visiting scholar at the Institute of Energy Technology, Aalborg University, Aalborg, Denmark. Dr. Gao was the recipient of the IEEE Industry Applications Society Industrial Power Converter Committee Prize for a paper published in 2006.

Peng Wang (M00) received the B.Sc. degree from Xian Jiaotong University, Xian, China, in 1978, the M.Sc. degree from Taiyuan University of Technology, Taiyuan, China, in 1987, and the M.Sc. and Ph.D. degrees from the University of Saskatchewan, Saskatoon, SK, Canada, in 1995 and 1998, respectively. Currently, he is an associate professor of Nanyang Technological University, Singapore.

Fook Hoong Choo received the B.Sc. degree from University of Leeds, Leeds, U.K., in 1977, and the M.Sc. degree from the University of Manchester, Manchester, U.K., in 1979. He was employed as a Design Engineer with GEC at Rugby, U.K. from 1979 to 1983 and Project Engineer with Lucas Research at Birmingham, U.K. from 1983 to 1984. He joined Nanyang Technological University, Singapore, (formerly Nanyang Technological Institute), in 1984 where he is currently an Associate Professor with the School of Electrical and Electronic Engineering. His current research interests include power electronics, ac drives, magnetics, renewable energy generation and control, and energy, water and environmental research.

Frede Blaabjerg (S86M88SM97F03) received the M.Sc. degree in electrical engineering from Aalborg University, Aalborg, Denmark, in 1987 and the Ph.D. degree from the Institute of Energy Technology, Aalborg University, in 1995. From 1987 to 1988, he was with ABB-Scandia, Randers, Denmark. In 1992, he became an Assistant Professor with Aalborg University, where in 1996, he became an Associate Professor and, in 1998, a Full Professor of power electronics and drives. In the period of 20062010, he was the Dean of the faculty of Engineering, Science, and Medicine at Aalborg University, Denmark. During the last years, he has held a number of Chairman positions in research policy and research funding bodies in Denmark. In 2007, he was appointed to the board of the Danish High Technology Foundation. He is the author or coauthor of more than 600 publications in his research elds, including the book Control in Power Electronics (Eds. M. P. Kazmierkowski, R. Krishnan, F. Blaabjerg) (Academic Press, 2002). His research areas are in power electronics, static power converters, ac drives, switched reluctance drives, modeling, characterization of power semiconductor devices and simulation, power quality, wind turbines, custom power systems, and green power inverter. Dr. Blaabjerg has been an Associate Editor of the IEEE T RANSACTIONS ON I NDUSTRY A PPLICATIONS , IEEE T RANSACTIONS ON P OWER E LEC TRONICS , Journal of Power Electronics, and of the Danish journal Elteknik. In 2006, he was the Editor-in-Chief of the IEEE T RANSACTIONS ON P OWER E LECTRONICS. He was the recipient of the 1995 Angelos Award for his contribution in modulation technique and control of electric drives and an Annual Teacher prize from Aalborg University in 1995. In 1998, he was the recipient of the Outstanding Young Power Electronics Engineer Award from the IEEE Power Electronics Society. He was also the recipient of nine IEEE Prize Paper Awards during the last ten years, the C. Y. OConnor Fellowship in 2002 from Perth, Australia, the Statoil Prize in 2003 for his contributions in power electronics, and the Grundfos Prize in 2004 for his contributions in power electronics and drives. From 2005 to 2007, he was a Distinguished Lecturer for the IEEE Power Electronics Society. It is followed up as Distinguished Lecturer for the IEEE Industry Applications Society from 2010 to 2011.