Tracking harmonic current quickly and precisely is one of the keys to designing active power filters (APF). In the past, the current state feedback decoupling PI control was an effective means for threephase systems in the current control of constant voltage constant frequency inverters and high frequency PWM reversible rectifiers. This paper analyzes in detail the limitation of the conventional PI conditioner in the APF application field and presents a novel PI control method. Canceling the delay of one sampling period and the misadjustment for tracking the harmonic current is the key problem of this PI control. In this PI control, the predictive output current value is obtained by a state observer. The delay of one sampling period is remedied in this digital control system by the state observer. The predictive harmonic command current value is obtained by a repetitive predictor synchronously. The repetitive predictor can achieve better predictions of the harmonic current. By this means, the misadjustment of the conventional PI control for tracking the harmonic current is cancelled. The experiment results with a threelevel NPC APF indicate that the steadystate accuracy and dynamic response of this method are satisfying when the proposed control scheme is implemented.
I. INTRODUCTION
Harmonics cause serious problems for power conversion systems. Passive filters that consist of
L
and
C
are generally implemented to attenuate the harmonics generated in the power system. However, passive filters have disadvantages in that the source impedance affects the filtering characteristics and amplification of the harmonic currents caused by the parallel resonance. The control of harmonic perturbations by active power filters (APF) has become a hot topic in the power engineering field. In the past two decades, active power filters have been developed rapidly to suppress the harmonics in power distribution systems of 380V grids
[1]

[4]
. However, the voltage grids are 600V in many power distribution systems for making petroleum holes. 690V grids are widely used in many of the power distribution systems of mining. There are many voltage grids ranging from 600V to 1000V in steelmaking factories. Due to the limitations of the voltage and current capabilities of power devices, it is very difficult to handle the nonlinear loads in the above mentioned voltage grids with the traditional APFs of a 380V grid with a twolevel inverter. In recent years, threelevel neutralpointclamped (NPC) PWM inverters have become an effective and practical solution for many moderate voltage application fields
[5]
. Evidently, the threelevel inverter is very suitable for the harmonic restriction of the above mentioned voltage grids. By adopting this topology, a threelevel APF with a 1200V IGBT may be used directly in 600V and 690V voltage grids, and a threelevel APF with a 1700V IGBT can be applied directly to 1000V voltage grids. Moreover, this topology can synthesize more output levels which is good for harmonic current tracking. In addition, the threelevel inverter can be combined with a passive filter to form the hybrid filter used in 6.6kV grids. The hybrid filter can use a 1200V IGBT at a reasonable cost from the market
[6]
,
[7]
. The APFs with a threelevel NPC inverter have received more and more attention in recent years
[6]

[12]
.
The high quality system current waveforms of a power system evaluated with little total harmonic distortion (THD) are a basic requirement for APFs
[13]

[21]
. Among them, the current state feedback decoupling PI control is distinguished by its simple structure and fast response while this control scheme provides lower current ripples and a constant switching frequency. It is widely applied in the current waveform control of threephase inverters, PWM rectifiers, STATCOMs and motor drives with sinusoidal inputs or outputs
[13]

[18]
. However, when this control is applied to an APF, there are some new problems that need to be solved. Firstly, the PI controller may track the DC signal with zero steadystate error. When the fundamental signal is converted into a DC quantity through synchronous rotating transformation in threephase inverters, STATCOMs and PWM rectifiers, the conventional PI controller may track the command signal in the rotating
dq
frame with zero steadystate error. However, the command current of the APF includes a large number of harmonics. The bandwidth of the conventional PI controller is almost enough for effectively tracking the command harmonic current. As a result, the conventional PI controller will have a large tracking error and the system current waveform filtered has a large THD. How to eliminate the tracking error?
The execution of the control program needs time. The calculation of the command voltage should be carried out a sampling period ahead, otherwise the maximum output pulse width will be significantly reduced. Under most conditions, the calculation of the command voltage in this sampling period will be carried out in the next sampling period. Therefore, a delay of one sampling period will occur. When the frequency of the command signal is comparatively low in threephase inverters, STATCOMs and PWM rectifiers, the tracking error is very small and may be neglected. However, the frequency of the harmonic current compensated by the APFs is comparatively high. If the delay is not remedied, the delay of one sampling and calculation period will seriously affect the filter effect of this APF. Therefore, the delay of one sampling period must be remedied.
The authors of
[22]
present a multiple synchronous rotating reference frame PI controller for active filters. Based on this feature, the preselected undesired harmonic currents are independently compensated with PI regulators in multiple synchronous reference frames. The limitation of this method consists of the significant computational burden arising from the need for multiple synchronous rotating reference frames. Moreover the number of reference frames will be doubled under unbalanced conditions. In
[23]
, a PIS control method based on adding an S (sine) regulator to a conventional PI regulator was presented for APF applications. The control method includes seven S regulators having different resonant frequencies. Each of the resonant frequencies is tuned to the fundamental, 5th, 7th, 11th, 13th, 17th, and 19th harmonics. An experiment carried out using a 1kVA APF prototype indicates that the control method has perfect currenttracking performance. However, this control method requires many S regulators. As a result, its structure is very complicated. Moreover, when the characteristics of the load current change and the new harmonic components are produced, the control effects of this method will deteriorate. The authors of
[24]
presented a fuzzy proportionalintegral controller (FPIC) used to provide the appropriate switching pattern of an APF to generate the actual compensation current. However, the THDs of the compensated experimental line currents were somewhat larger and these experimental results cannot agree reasonably well with the theoretical predictions.
This paper analyzes the shortages of the conventional current state feedback decoupling PI control in detail and concludes that canceling the delay of one sampling period and remedying the misadjustment for tracking the harmonic current are the key problems of this PI control method. Then a novel PI control method is presented. In this PI control, the predictive output current value is obtained by a state observer. The delay of one sampling period in this digital control system is remedied by the state observer. The predictive harmonic command current value is obtained by a repetitive predictor synchronously. The repetitive predictor can achieve better prediction of the harmonic current. By this means, the misadjustment of the conventional PI control for tracking the harmonic current is cancelled. This PI control method has a fast dynamic convergence rate and little static misadjustment. The experiment results illustrate that the performance of the proposed approach is satisfactory.
II. OPERATING PRINCIPLE AND CONTROL SYSTEM
The operating principle of the APF is shown in
Fig.1
. Afterthe load currents flow through the current transformers, the desired currents, which the APF should compensate, are first obtained by the detecting circuit of the harmonic currents. Then the desired currents are injected into the power system by the APF controlling the PWM inverter. It can be seen from
Fig. 1
that the source current
i_{s}
is provided with both the load current
i_{l}
and the output current
i_{c}
of the APF, i.e.,
i_{s}
=
i_{l}
+
i_{c}
. Since
i_{c}
= 
i_{h}
, the power source only supplies
i_{1}
. Hence,
i_{s}
, after compensation, becomes a sinusoidal current which has a frequency that is identical to
u_{s}
. This is the basic operating principle of the APF. The threelevel NPC inverter applied to the APFs, as shown in
Fig. 1
, is builtup of twelve switches, each with a freewheeling diode, and six power diodes. This allows for the connection of the phase outputs to the middle voltage. The blocking voltage of each switch is 1/2 of the DC voltage. Each arm of the inverter can be clamped to the DC terminals P, N and Q, to produce three switching states.
Schematic diagram of the active power filter with threelevel inverter.
When the upper two switches (T
_{1}
and T
_{2}
) are switched on, the output of this phase is connected to the DC terminal P. When the middle two switches (T
_{2}
and T
_{3}
) are switched on, the output of this phase is connected to the middle point N. Similarly when the lower two switches (T
_{3}
and T
_{4}
) are switched on, the output of this phase is connected to Q. The allowed logic configurations of the NPC switches are able to provide the three different output voltage values furnished by each NPC inverter phase. The states of T
_{1}
, T
_{2}
, T
_{3}
and T
_{4}
are complementary. The two switches for each phase of the NPC inverter are closed, whilst the other two are opened at every time instant
[5]
.
From
Fig.1
, the control system model for tracking the command current of the threelevel APF in the stator coordinates is established as follows:
In this paper, the power loss on the resistor R is assumed to include all of the filter's losses. Where
u_{sa}
,
u_{sb}
and
u_{sc}
are the utility voltages,
u_{a}
,
u_{b}
and
u_{c}
are output voltages of the threelevel inverter. Through a synchronous rotating transformation, the equation in the ABC stationary frame can be expressed in the
dq
frame as follows:
By this means, the threephase alternating signal is changed into a twophase alternating signal. Thus, the mathematic model of this APF is predigested. A diagram of the control system is shown in
Fig.2
. The control system includes an outerloop voltage controller and an innerloop current controller. The outerloop controller is used to produce the command active current to regulate the DC voltage. The innerloop current controller is divided into the detection module for calculating the harmonic current and the control module for tracking the command current and restraining the neutralpoint voltage imbalance. In order to control the shape of the output current waveform, a novel PI control technology is presented. The neutralpoint voltage imbalance is restrained by selecting the small vectors that will move the neutralpoint voltage in the direction opposite the direction of the unbalance from the space vector PWM point of view in the control module
[25]
.
Block diagram of control system of the active power filter with threelevel inverter.
III. NOVEL PI CONTROL METHOD
Traditionally, the current state feedback decoupling PI control is an effective means for threephase systems in the current control of high frequency PWM reversible rectifiers. From (2), a block diagram of the PI control system is as shown in
Fig. 3
[13]

[15]
,
[20]
. Through this PI control method, the equivalent decoupling control diagrams for
i_{d}
and
i_{q}
can be derived as shown in
Fig. 4
.
KP
and
KI
are the proportional and integral parameters of the PI controller, respectively. The control plant is a firstorder. The current control closedloop with the PI controller is a typical secondorder system.
Block diagram of current state feedback decoupling PI control.
Block diagram of equivalent decoupling control.
The current control closedloop is a typical secondorder system which can be corrected by a PI controller into a firstorder element by means of the Siemens method. The Siemens method is widely used in the actual tuning of the PI parameters. The core thought of this design is to let the zero of the controller offset the pole of the plant in the
s
plane. That is to say,
K_{P}
/
K_{I}
=
L_{s}
/
R_{s}
. Using this method, the PI parameter
K_{P}
and
K_{I}
are designed as follows:
Where
T_{s}
is sampling\switch period. Substituting equation (4) into (3) yields the following results:
From expressions (5), it can be seen that the closedloop current control system is corrected into a firstorder element. The time constant
T_{P}
of the closedloop current control is equal to the switch period
T_{s}
.
The bode map of
C(s)
is as
Fig. 5
. From the bode map of
C(s)
,
C(s)
has a closetounit gain in the low frequency band. Because the fundamental signal is converted into a DC quantity through a synchronous rotating transformation, the conventional PI controller, which is designed with abovementioned method, has little static misadjustment in the PWM reversible rectifier. However, the command currents of the APFs include a large number of harmonics. From the bode map of
C(s)
, the magnitude of
C(s)
degrades in the high and medium frequency bands. The bandwidth of this PI controller is almost enough for effectively tracking the command harmonic current. The PI controller has a static misadjustment for tracking the command current, and the system current waveform filtered has a large THD. In order to make the output current track the compensation harmonic reference well, the gain of the PI controller should be increased. However, doing that might cause oscillations and instability of the APF.
bode map of C(s).
From expressions (5), the discrete equivalent equation of the closedloop current control system can be represented as:
Simplifying expressions (6), the answer can be represented as:
From expressions (7), it can be seen that the output compensation current at the
(k+1)th
sampling instant is equal to the command harmonic current at the
kth
instant in the abovementioned digital PI control system. Therefore, the output current lags the command current by one sampling period. Thus, the controller has a static misadjustment. If the command harmonic currents
i
^{*}
_{c}
(
k
+ 1) can be predicted at the
kth
instant and
i
^{*}
_{c}
(
k
+ 1) is substituted for
i
^{*}
_{c}
(
k
) in equation (7), the following results may be obtained:
By this means, the misadjustment of the conventional PI control for tracking the harmonic current is cancelled. Therefore, predicting the harmonic currents at the coming sampling instant precisely is an effective method to improve this controller.
Moreover, the execution of the control program needs time. The calculation of the command voltage should be carried out a sampling period ahead, otherwise the maximum output pulse width will be significantly reduced. When the calculation of the command voltage is carried out in the above sampling period, that is to say, the command voltage calculated in the
(k)th
sampling period is carried out in the
(k+1)th
sampling period, a delay of one sampling period will occur. The question can also be represented as:
If the control system follows the command current accurately, the output current of the APF can be represented as:
Expression (10) is the same as expressions (7). Therefore, the bode map of expression (10) is the same as
Fig.5
. The frequency of the harmonic current compensated by the APF is higher than the inverter, the STATCOM and the PWM rectifier. From the bode map in
Fig.5
, the delay error of one sampling period in the APF system (when the harmonic required to be compensated is 3KHz) is large. Furthermore, if the delay of one sampling and calculation period is not remedied, the delay will seriously affect the filter effect of this APF. As a result, remedying the delay of one sampling period is another key problem in improving this controller.
Therefore, the control process should be like this: at the
(k1)th
instant, sample the output current
i_{c}(k1)
of the APF, then estimate
i_{c}(k)
. Meanwhile, sample the load current
i_{l}(k1)
and calculate the harmonic current
i_{h}(k1)
, then predict
i_{h}(k+1)
. Finally the predicted value
i_{h}(k+1)
is put into the PI controller of the closedloop current control system. Using this approach, the control system can track the command current with no static misadjustment.
A good solution for the estimation of the APF output currents
i_{c}(k)
is a state observer. From (2), the control system model for tracking the command current can be changed as follows:
Then the sample data model becomes as follows:
A full order state observer for this APF is as follows:
Where
T
is the observer gain. The state predictor is as
Fig.6
.
denote the estimated values of the output currents at the
kth
and
(k+1)th
sampling instants.
The state predictor.
The observer gain is decided so that a stable pole placement is realized in the observer.
From (12) and (13), the error system is:
Where
is the observer error. The poles of the observer can be arbitrarily placed by adjusting the feedback gain
T
. In practice, it is easy to make the observer error quite close to zero by setting the poles of the observer suitably.
It seems impossible to model the load current disturbance precisely when considering the diversity of load types. Since the harmonic is repetitive most of the time, it makes sense to adopt a repetitive algorithm to predict the seemingly unpredictable load currents
[20]
. The basic theorem of the repetitive harmonic predictor is show as
Fig.7
.
The repetitive predictor.
The algorithm can be seen as a plugin repetitive compensator for a basic predictor. The basic predictor is a unit delay: z
^{2}
, which means that it just takes the value of harmonic current at the
kth
sampling instant (namely
i_{hd}(k)
,
i_{hq}(k)
) for the harmonic current values at the
(k+2)th
sampling instant since the
(k+2)th
sampling process has not yet begun. The input of the compensator is
e
, which is defined as the difference between the value of the actual current and the estimated value. It is assumed that harmonic current is sampled
N
times in a fundamental cycle. In order to repetitively compensate the error of the basic prediction,
N
memory cells are prepared to store the correction values (namely
△(k)
,
k∈[1,N]
) in one fundamental cycle. The predicted value of the harmonic current at the
(k+2)th
sampling instant (namely
) can be obtained by adding
△(k)
to the value of the harmonic current at the kth sampling instant, as (13). The correction value
△(k)
for correcting the basic prediction value
i_{h}(k)
to predict
is modified periodically by adding the prediction error to the current correction value
△(k)
. The prediction error is obtained by subtracting the predicted value of the harmonic current
from the actual harmonic current value
i_{h}(k+2)
.
Due to the unperfected modeling, a bandlimit filter is often included in a repetitive controller. Such a filter can improve the stability of the filter system by eliminating the error
e
by half. This filter equals 0.95 (the parameter was shown to be suitable in most of our simulations). In this case, the basic estimator is just a simple and fixed algorithm, and not a real physical system where measuring errors and parameter drift are inevitable. As for the gain
k_{r}
, when it is set to be unity, the error will be compensated nearly 100% in the second cycle under a repetitive input. However, in practice, the load current does not exactly repeat itself from cycle to cycle. For a smooth convergence of the error,
k_{r}
should be chosen as less than 1. The simulation results show that the value
k_{r}
=0.98 appear to be a good choice for all of the experiments.
IV. EXPERIMENTAL RESULTS
The experimental parameters are shown as follows. The nonlinear load is a threephase bridge rectifier with inductance and resistance. The source RMS phase voltage value is 110V and the power frequency is 50Hz. The impedance of the voltage source is 1mH. The filter inductance is 2mH and the resistor of the APF output filter is 0.5Ω. The DC voltage of the filter is 360V and the DC capacitor of the filter is 4700μF. Both the sampling frequency of the control system and the switch frequency of the switching device are 9.6kHz. The DC controller parameter kp is 1.6 and kI is 64. Repetitive predictor parameter kr is 0.98. A PMOSFET 2SK1020 is selected as the switching device. Some important experimental oscillograms are shown as follows:
Fig. 8
shows the waveforms of both the load and utility current through the APF filtering. Waveform 2 is the utility current and waveform 1 is the load current with the help of the harmonic compensation of the threelevel APF with the new PI controller. It can be calculated from these figures that the value of the total harmonic distortion (THD) can be reduced from 22.54% to 3.3%.
Fig.9
shows the waveforms of both the load and utility current through the APF filtering with the traditional PI controller. The THD value of the utility current can be calculated as 6.1%. Obviously, the filtering ability of this new PI controller is better than the traditional PI controller.
Fig.8

9
prove that the new proposed PI controller has good steady filtering capability.
Experimental steady waveforms o f load current and system supply current.
Experimental steady waveforms of lo ad current and system supply current with this traditional PI controller.
Experimental dynamic waveforms of load current and system supply current when nonlinear load is added abruptly.
In order to test the dynamic response characteristic of this PI controller, some of the dynamic waveforms were recorded when load switched on and off, as shown in
Fig.10
~
11
respectively. As shown in
Fig.10
, from waveform 2, it can be seen that the composite controller responds to the current reference very quickly. When the load is switched on suddenly, the reference distortion lasts only two periods. In
Fig.11
, waveform 2 shows the output current of the APF when the load is switched off. The PI controller also responses very quickly. It is obvious that the proposed APF is controlled very well by the PI control.
Experimental dynamic waveforms of load current and system supply current when nonlinear load is removed abruptly.
Experimental waveforms of u_{AN} and u_{AB}.
The experimental waveforms of
u_{AN}
and
u_{AB}
are illustrated in
Fig.12
, where waveform 1 is
u_{AN}
and waveform 2 is
u_{AB}
. It is obvious that the output voltage
u_{AN}
furnished by each NPC inverter phase is threelevel and that the lineline output voltage
u_{AB}
is fivelevel.
Fig.12
shows that the proposed APF is suitable for the application field of voltage grids ranging from 600V to 1000V.
V. CONCLUSIONS
Based on a detailed performance analysis of the traditional current state feedback decoupling PI control for APFs, this paper proposed a novel PI controller to improve the steady state compensation precision and ensure the dynamic response performance. This novel PI control algorithm can suppress the periodic error in the whole system to achieve zero steady error tracking. The feasibility of this control method is verified by theoretical analysis and laboratory tests.
Acknowledgements
This work was supported by Innovation Foundation of State Key Laboratory of Electrical Insulation and Power Equipment (No.EIPE13307), Delta Educational Development Foundation (DREG2013007) and with grants from National High Technology Research and Development Program of China (2012AA050206) and National Science Foundation of China (No.50907052).
BIO
Yingjie He was born in Henan Province, China, in 1978. He received his B.S., M.S. and Ph.D. from the Huazhong University of Science and Technology, Wuhan, China, in 1999, 2003 and 2007, respectively. From May 2007 to May 2009, he was with the Power Electronics and Renewable Energy Center at Xi’an Jiaotong University, Xi’an, China, as a Postdoctoral Research Scholar. He is currently a Lecturer at Xi’an Jiaotong University. His current research interests include power quality control, multilevel inverters and the application of power electronics in power systems.
Jinjun Liu was born in Hunan Province, China, in 1970. He received his B.S. and Ph.D. from Xi’an Jiaotong University (XJTU), Xi’an, China, in 1992 and 1997, respectively. In 1998, he led the founding of the XJTU/Rockwell Automation Laboratory. From December 1999 to February 2002, he was with the Center for Power Electronics Systems at the Virginia Polytechnic Institute and State University, Blacksburg, VA, USA, as a Postdoctoral Research Scholar. After that he came back to XJTU and in August of 2002 was promoted to Full Professor and Head of the Power Electronics and Renewable Energy Center. He is also serving as an Associate Dean in the School of Electrical Engineering at XJTU. He has coauthored 3 books, published over 100 technical papers, and received several provincial or ministerial awards for scientific and career achievements. He was also a recipient of the 2006 Delta Scholar Award. His current research interests include power quality control, renewable energy generation, utility applications of power electronics, and the modeling and control of power electronic systems. Dr. Liu has served as the IEEE Power Electronics Society Region 10 Liaison for 3 years. He was actively involved in the organization of several power electronic international conferences, including PESC, APEC, IPEC in Japan, ICPE in Korea, and IPEMC in China, where he served as a Committee Member, a CoChair, or a Session Chair.
Zhaoan Wang was born in Shaanxi Province, China, in 1945. He received his B.S. and M.S. from Xi’an Jiaotong University, Xi’an, China, in 1970 and 1982, respectively, and his Ph.D. from Osaka University, Osaka, Japan, in 1989. From 1970 to 1979, he was an Engineer at the Xi’an Rectifier Factory, Xi’an, China. He is now a Professor at Xi’an Jiaotong University. His current research interests include power conversion systems, harmonics suppression, reactive power compensation and active power filters.
Yunping Zou was born in Hunan Province, China, in 1945. He received his B.S. from the Huazhong University of Science and Technology, Wuhan, China, in 1969. He has been a Full Professor at the Huazhong University of Science and Technology since 1994. His current research interests include new power electronic circuits, devices and systems; digital intelligent control technology and its basic application; and signal detection, transforming and processing.
Kumar P.
,
Mahajan A.
2009
“Soft computing techniques for the control of an active power filter,”
IEEE Trans. Power Del.
24
(1)
452 
461
DOI : 10.1109/TPWRD.2008.2005881
RamosCarranza H. A.
,
Medina A.
,
Chang G. W.
2008
“RealTime Shunt Active Power Filter Compensation,”
IEEE Trans. Power Del.
23
(4)
2623 
2625
DOI : 10.1109/TPWRD.2008.2002836
Singh B.
,
Verma V.
2008
“Selective compensation of powerquality problems through active power filter by current decomposition,”
IEEE Trans. Power Del.
23
(2)
792 
799
DOI : 10.1109/TPWRD.2007.911108
Hirve S.
,
Chatterjee K.
,
Fernandes B. G.
,
Imayavaramban M.
,
Dwari S.
2007
“PLLLess Active Power Filter Based on OneCycle Control for Compensating Unbalanced Loads in ThreePhase FourWire System,”
IEEE Trans. Power Del.
22
(4)
2457 
2465
DOI : 10.1109/TPWRD.2007.893450
Akira N.
,
Takahashi I.
,
Akagi H.
2007
“A new neutralpointclamped PWM inverter,”
IEEE Trans. Ind. Applicat.
17
(3)
518 
523
Akagi H.
,
Kondo R.
2009
“A transformerless hybrid active filter using a threelevel PWM converter for a mediumvoltage motor drive,”
in Proc. ECCE
1732 
1739
Akagi H.
,
Hatada T.
2009
“Voltage Balancing Control for a ThreeLevel DiodeClamped Converter in a MediumVoltage Transformerless Hybrid Active Filter,”
IEEE Trans. Power Del.
24
(3)
571 
579
DOI : 10.1109/TPEL.2009.2012528
Wong M. C.
,
Tang J.
,
Han Y. D.
2003
“Cylindrical coordinate control of 3D PWM technology in 3phase 4wire trilevel inverter,”
IEEE Trans. Power Electron.
18
(1)
208 
220
DOI : 10.1109/TPEL.2002.807133
Dai N. Y.
,
Wong M. C.
,
Han Y. D.
2006
“Application of a threelevel NPC inverter as a threephase fourwire power quality compensator by generalized 3DSVM,”
IEEE Trans. Power Electron.
21
(2)
440 
449
DOI : 10.1109/TPEL.2005.869755
Basu M.
,
Das S. P.
,
Dubey G. K.
2004
“Parallel converter scheme for highpower active power filters,”
IEE Proceedings Electric Power Applications
151
(4)
460 
466
DOI : 10.1049/ipepa:20040211
Vodyakho O.
,
Mi C. C.
2009
“Threelevel inverterbased shunt active power filter in threephase threewire and fourwire systems,”
IEEE Trans. Power Electron.
24
(5)
1350 
1363
DOI : 10.1109/TPEL.2009.2016663
Akagi H.
,
Inoue S.
,
Yoshii T.
2007
“Control and performance of a transformerless cascade PWM STATCOM with star configuration,”
IEEE Trans. Ind. Applicat.
43
(4)
1041 
1049
DOI : 10.1109/TIA.2007.900487
Maharjan L.
,
Inoue S.
,
Akagi H.
2008
“A transformerless energy storage system based on a cascade multilevel PWM converter with star configuration,”
IEEE Trans. Ind. Applicat.
44
(5)
1621 
1630
DOI : 10.1109/TIA.2008.2002180
Akagi H.
,
Fujita H.
,
Yonetani S.
,
Kondo Y.
2008
“A 6.6kV transformerless STATCOM based on a fivelevel diodeclamped PWM converter: System design and experimentation of a 200V 10kVA laboratory model,”
IEEE Trans. Ind. Applicat.
44
(2)
672 
680
DOI : 10.1109/TIA.2008.916713
Zhou Z.
,
Unsworth P. J.
,
Holland P. M.
,
Igic P.
2009
“Design and analysis of a feedforward control scheme for a threephase voltage source pulse width modulation rectifier using sensorless load current signal,”
IET Proceedings Power Electronics
2
(4)
421 
430
DOI : 10.1049/ietpel.2008.0058
Lee K.
,
Jahns T. M.
,
Lipo T. A.
,
Blasko V.
,
Lorenz R. D.
2009
“Observerbased control methods for combined sourcevoltage harmonics and unbalance disturbances in PWM voltagesource converters,”
IEEE Trans. Ind. Applicat.
45
(6)
2010 
2021
DOI : 10.1109/TIA.2009.2031854
Yoo A.
,
Sul S. K.
2009
“Design of flux observer robust to interior permanentmagnet synchronous motor flux variation,”
IEEE Trans. Ind. Applicat.
45
(5)
1670 
1677
DOI : 10.1109/TIA.2009.2027516
Fujita H.
2009
“A singlephase active filter using an hbridge PWM converter with a sampling frequency quadruple of the switching frequency,”
IEEE Trans. Power Electron.
24
(4)
934 
941
DOI : 10.1109/TPEL.2008.2009302
Zhang K.
,
Kang Y.
,
Xiong J.
,
Chen J.
1999
“Deadbeat control of PWM inverter with repetitive disturbance prediction,”
in Proc. APEC
1026 
1031
Vodyakho O.
,
Hackstein D.
,
Steimel A.
,
Kim T.
2008
“Novel direct currentspacevector control for shunt active power filters based on the threelevel inverter,”
IEEE Trans. Power Electron.
23
(4)
1668 
1678
DOI : 10.1109/TPEL.2008.925181
Bojrup M.
,
Karlsson P.
,
Alakula M.
,
Gertmar L.
1999
“A multiple rotating integrator controller for active filters,”
in Proc. EPE.
756 
762
Fukuda S.
,
Imamura R.
2005
“Application of a sinusoidal internal model to current control of threephase utilityinterface converters,”
IEEE Trans. Ind. Electron.
50
(2)
420 
426
DOI : 10.1109/TIE.2005.843914
Kirawanich P.
,
O’Connell R. M.
2004
“Fuzzy logic control of an active power line conditioner,”
IEEE Trans. Power Electron.
19
(6)
1574 
1585
DOI : 10.1109/TPEL.2004.836631
Celanovic N.
,
Voroyevich D.
2000
“A comprehensive study of neutralpoint voltage balancing problem in threelevel neutralpointclamped voltage source PWM inverters,”
IEEE Trans. Power Electron.
15
(2)
242 
249
DOI : 10.1109/63.838096