Electrical Power and Energy Systems 105 (2019) 717–728

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

A controllable inductive power filtering system: modeling, analysis and T

control design☆
⁎
Qianyi Liu, Yong Li , Sijia Hu, Longfu Luo
College of Electrical and Information Engineering, Hunan University, Changsha 410082, China

A R T I C LE I N FO A B S T R A C T

Keywords: This paper presents a controllable inductive power filtering (CIPF) system with the novel filtering mechanism.
Power quality Due to the zero-impedance design of the power transformer, the harmonic magnetic flux between the secondary
Inductive power filtering load winding and the customized filtering winding can be balanced each other. Hence, the contaminated area of
Transformer harmonics is reduced. The addition of the active filter contributes to further improve the filtering performance
Harmonic suppression
with the task of harmonic isolation between the power grid and the load. First, the three-phase unified
equivalent circuit model and the mathematical model are given out at the harmonic domain. Based on the
deduced transfer matrix, the filtering evaluation criterion of CIPF system is established, and the filtering prin-
ciple is revealed, as well. According to the special operational law, an improved virtual impedance control
strategy is designed to prevent the potential harmonic amplification, and the relationship between the phase/
magnitude of the virtual impedance and the filtering performance is discussed. Finally, the experimental results
based on the laboratory scaled prototype show the validity of the CIPF system with improved virtual impedance
control.

1. Introduction compensation range of reactive power can be dynamically regulated

Rapidly expanding applications of the power electronic devices in-
evitably introduce power quality (PQ) issues to the public grid [1–4].
Harmonic pollution and reactive power loss are the two main PQ issues
existed in the three-phase power system. Harmonic may result in the
malfunction of the sensitive equipment nearby the harmonic source [5].
The operating loss caused by these undesirable roving components on
the electrical equipment is non-negligible, as well. For instance, the
transformer may need to be derated as much as 50% capacity in the
environment with the extremely distorted current waveform and low
power factor [6,7]. The temperature rise due to harmonic contamina-
tion accelerates the insulation aging of the transformer, which shortens
its service life [8].
To improve the power supply quality, the tuned filter (L-C) with low
cost and simple structure is widely adopted to suppress harmonics and
provide constant reactive power compensation in the early time [9].
Static var compensator (SVC) is later used in the industrial applications
with the topology of thyristor controlled reactor (TCR), thyristor swit-
ched capacitor (TSC) or thyristor controlled LC (TCLC), whose
according to the load variation [10–12]. However, SVC has the dis-
advantage of undesirable harmonic emission. More importantly, both
tuned filter and SVC have the risk of parallel/series resonance with the
grid impedance [13]. With the remarkable progress in semiconductor
switching devices, the research hotspot shifts to the active filtering
method, which is able to overcome these negative effects [14]. Active
power filter (APF) is more effective in harmonic elimination compared
with passive filter. But high-power converter rating is required when
used in the large power system, which inevitably makes the cost of APF
quite high. The main connection schemes of APF are summarized as
shown in Fig. 1 [15–25]. In type I, an additional step-down transformer
is used to reduce the voltage level [15,16]. The auxiliary transformer
accounts for around one-third of total weight, thus makes the filter
bulky and heavy [17]. Literatures [17] and [18] present the PQ con-
ditioner with cascade multi-level converter (CMC) to realize a small and
lightweight filter in middle voltage (MV) or high voltage (HV) appli-
cations, similar like type II. In practice, the hard selection between the
cascaded count and the specification of CMC module determines the
cost and performance of the APF. Type III is the hybrid APF (HAPF),

☆
This work was supported in part by the 111 Project of China under Grant B17016, in part by the Huxiang Youth Talent Program of Hunan Province under Grant
2015RS4022, in part by the Excellent Innovation Youth Program of Changsha of China under Grant KQ1707003, and in part by the Key S&T Special Project of Hunan
Province of China under Grant 2017GK2241.
⁎
Corresponding author.
E-mail addresses: liu7y@foxmail.com (Q. Liu), yongli@hnu.edu.cn (Y. Li), huda_hsj@163.com (S. Hu), llf@hnu.edu.cn (L. Luo).

https://doi.org/10.1016/j.ijepes.2018.09.016
Received 10 June 2018; Received in revised form 28 July 2018; Accepted 8 September 2018
Available online 20 September 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
Q. Liu et al.
Fig. 1. Main connection schemes of APF.
which includes the structures of LC + APF [19–21], TCR + APF, [22],
TCLC + APF, [23], etc. The main tasks for harmonic suppression and
reactive power compensation are undertaken by LC or SVC, while the
active part is controllable to enhance the filtering performance of the
HAPF. In type III, most of the fundamental voltages drop on the passive
components, so that the operating capacity of the inverter is greatly
reduced. For insulation consideration, the rating voltage of the active
filter should still be designed according to the voltage level at the access
point (ac bus) [21]. The rated voltage of the active component in type
III is still high in practice. Types IV and V which use the power trans-
former as the auxiliary transformer simultaneously are further derived.
Type IV is called as winding-taps-injection DSTATCOM, which is con-
nected to the taps of the primary winding [24]. The topology is easy to
realize and expand. However, the compensation performance is ser-
iously affected in the case of high load ratio.
In some high-power industrial applications, it is impossible for the
filter to install at secondary side of the transformer due to the existence
of large current. However, the undesirable components, in which the
harmonic and reactive power are the dominant contaminant, flowing
through the transformer make the operating loss high. To solve such
issue, the power transformer is redesigned in [25], and a controllable
inductive power filter (CIPF) using the customized filtering winding is
presented as shown in type V. The equivalent impedance of the filtering
winding is zero. At specific harmonic frequency, the harmonics can be
canceled out between the load winding and the filtering winding, so
that no harmonic is transferred into the grid winding [26]. In this way,
the CIPF system can weaken the negative effect of harmonic and re-
active components on the transformer. Moreover, a compromise be-
tween the voltage rating of the inverter and the initial investment is
also realized by means of the filtering winding. Literature [25] proposes
a comprehensive virtual impedance control to realize the zero-im-
pedance design for implementing the inductive filtering. Furthermore,
the feasibility for the CIPF applied in the shipboard power system is
discussed in [27].
Totally different from the existing literatures, to further explore the
filtering performance of the CIPF system, this paper mainly focuses on
the following aspects.
• Establishing the three-phase equivalent mathematical model for
CIPF system, which includes the transfer relationship between the
grid-side current with the load harmonic current and the back-
ground harmonic voltage;
• Revealing the distinctive operational mechanism of the CIPF system
in a mathematical sense, based on the obtained transfer matrices;
• Finding out the operational law. An improved control strategy is
designed, and the optimal complex impedance is identified.
This paper is organized as follows. Section 2 describes the system
718
Electrical Power and Energy Systems 105 (2019) 717–728
topology and technical features of the DC supply system based on the
proposed CIPF method. The mathematical model is established in
Section 3. Section 4 analyzes the operational mechanism and the fil-
tering performance. Moreover, Section 5 presents an improved control
strategy, and the filtering ability is further investigated. Section 6 gives
the experimental results. Finally, the conclusion is given in Section 7.
2. System topology
Fig. 2 shows the new DC supply system based on the CIPF method.
Different from the conventional filter, CIPF (blue area) is installed be-
hind the PCC/primary bus. As the key equipment for implementing the
inductive power filtering (IPF) method, the new transformer, which has
a three-winding structure, is configured in the vicinity of the rectifier
bridge. Besides, its primary winding adopts the star wiring, and is
connected with the public power system (red area). The secondary
winding of the new transformer adopts the extended-delta wiring,
which not only provides a close delta loop for the zero-sequence cur-
rent, but also has the merits of high capacity utilization and easy to
realize phase shift. The extended winding connects to the rectifier load
(green area), and serves as the load winding. For each phase, there is a
linking point between the delta winding and the extended winding. The
three parallel filtering branches access to the new transformer via the
linking point. The main distinction between the IPF and the CIPF is the
configuration of filtering branches. In this paper, the proposed CIPF
uses a 5th single-tuned filter and a series voltage-source inverter (VSI)
as the filtering branches.
The proposed DC supply system has the following technical features:
(1) The filtering system takes the full advantage of IPF method to re-
duce the most effects of harmonic on the transformer. The flowing path
of the harmonic current is restricted in a small area, so that the purity of
the grid-side current waveform can be satisfied; (2) Thanks to that the
filtering winding can be customized for the practical application, the
controlled VSI can directly access to the transformer at an acceptable
voltage level without additional auxiliary transformer. (3) Hybrid filter
further cuts down the capacity of active filter dramatically by sharing
the main task of harmonic suppression with the passive filter. (4) With
the adoption of the virtual impedance control, it can make up the IPF’s
defects to withstand the external fluctuation and further mitigate the
impact of harmonics from the grid.
3. Mathematical modeling
3.1. Basic current transitive relation
Referring to Figs. 2 and 3, and according to the principle of ampere-
turns balance, the winding currents in the new transformer should sa-
tisfy the following relationship, that is
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Fig. 2. Topology of new DC supply system based on the CIPF method.

N1 I1K + N2 I2K + N3 I3k = 0
where I1K, I2K, I3k (K = A, B, C and k = a, b, c) are the currents in the
grid winding, load winding and filtering winding, respectively; Ni is the
corresponding number of turns (i = 1, 2, 3).
According to the Kirchhoff’s current law, the equations of the cur-
rents in the secondary winding can be obtained:
⎧ ILA = I3a + Ifa−I3c
ILB = I3b + Ifb−I3a
⎨ (2) When the filter is in operation, at the fundamental frequency, the
⎩ ILC = I3c + Ifc−I3b
where ILK and Ifk are the load currents and filtering branch currents,
respectively.
The above (1) and (2) combined with Fig. 2 can be used to derive
(1) the basic current transitive relation between the primary winding and
the secondary winding, as shown in (3).
From (3), we can obtain the following conclusions:
(1) The grid winding currents are composed of ILK and Ifk. In other
word, I1K are affected by the variable filtering branch currents. The
regulation of Ifk can be realized by adjusting the output voltage of
VSI appropriately, so that no harmonic current can flow through the
grid winding theoretically.
(2)
capacitive current generated by the LC filter and the inductive
current generated by the rectifier bridge can be offset each other,
hence the reactive power nearby compensation is realized.

Fig. 3. Equivalent circuit model of the secondary winding. (a) Ifa = Ifb = Ifc = 0; (b) ILA = ILB = ILC = 0.

719
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Fig. 4. Harmonic circuit model of the new transformer.

N
2 N
3
⎡ − ( N1 + 3N1 )
N3
0 ⎤ the single tuned filter; VSkn are the background harmonic voltages.
⎢  3N1
⎥ I According to the multi-winding transformer theory, one can obtain
⎡ I1A ⎤ ⎢ d1 ⏟ d2 ⎥ ⎡ LA ⎤ that,
⎢ I1B ⎥ = ⎢ N2 N3 N3 ⎥ ⎢ ILB ⎥
⎢ I1C ⎥ ⎢ 0 − ( N1
+ 3N1
) ⎥ ⎢ ILC ⎥
3N1
⎣ ⎦ ⎢ ⎣ ⎦
N ⎥
N1 N1
N3
0
N
− ( N2 + 3N3 ) ⎥ ⎧V1An− N3 V3an = Z1n I1An− N3 Z3n I3an
⎢ 3N1 ⎪
⎣ 1 1 ⎦
N1 N1
N 3 N V1Bn− N V3bn = Z1n I1Bn− N Z3n I3bn
⎡ 3N3 − 3N1 0 ⎤ ⎨ 3 3

⎢ 1 ⎥ ⎡ Ifa ⎤ ⎡ ILA ⎤ ⎡ Ifa ⎤ ⎪ V1Cn− N1 V3cn = Z1n I1Cn− N1 Z3n I3cn

+⎢ 0
N3 N
− 3N3 ⎥ ⎢ Ifb ⎥ = TL ⎢ ILB ⎥ + Tf ⎢ Ifb ⎥ ⎩ N3 N3 (8)
⎢ 3N1 1⎥
⎢ ⎥ ⎢ ILC ⎥ ⎢ Ifc ⎥
⎢− N3 N3 ⎥ ⎣ Ifc ⎦ ⎣ ⎦ ⎣ ⎦
0 where ViKn and Zin (i = 1, 2 and 3) are the voltage and the equivalent
⎣ 3N1 3N1 ⎦ (3)
impedance of the transformer windings, respectively.
The aforementioned current Eqs. (1), (2) and the voltage Eqs.
3.2. Mathematical model at harmonic domain (4)–(8) constitute together the mathematical model of the three-phase
CIPF at the harmonic domain. The mathematical model reflects the
Fig. 4 gives the harmonic circuit model of the CIPF. As shown in restriction relationship among the harmonic voltage, the harmonic
Fig. 4, the VSI can be equivalent to a harmonic current controlled current and the harmonic impedance. Based on this model, the transi-
voltage source [25], whose voltage is tive relation between ILn and VSn with ISn are given out in the following
VCk = Kn·ISKn , Kn > 0. (4) two cases.

where the gain of Kn is defined as the virtual resistive impedance with Case 1: Considering the impact of load current on the grid-side
the dimension of [Ω]; ISKn is the harmonic current at the grid side; the current, let VSn = 0. The harmonic current transfer matrix TI can be
subscript ‘n’ represents the n-order harmonic components of the vari- obtained in (9). Wherein, ZP1, ZP2, ZA1 and ZA2 are the characteristic
able, and similarly hereinafter. impedance, and decided by the ZSn, Z1n, Z3n, Zfn and Kn together, as
According to the Kirchhoff’s voltage law, the voltages on the 3 3
shown in (A. 1) in Appendix, and Δ = ZA1 −ZA2 .
transformer winding can be obtained as follows, Case 2: Considering the impact of background voltage on the grid-
side current, let ILn = 0. The harmonic voltage transfer matrix TV
⎧VSAn = ZSn ISAn + V1An can be obtained in (10).
VSBn = ZSn ISBn + V1Bn
⎨
⎩ VSCn = ZSn ISCn + V1Cn (5)
Correctness verification: As listed in Table 1, when there is no
filter in operation, i.e., Zfn and Kn satisfy that |Zfn| → ∞ and Kn = 0, the
⎧ Vabn = Vaon−Vbon
Vbcn = Vbon−Vcon matrix TI can be rewritten as the same form as TL in (3). Moreover, the
⎨ matrix TV will turn to be zero, that is, VSn has no influence on the grid-
⎩ Vcan = Vcon−Vaon (6)
side currents on the condition that no filter is in operation. The cor-
⎧ Vaon = Zfn Iaon + VCa rectness verification shows that the derived matrices (TI and TV) are
Vbon = Zfn Ibon + VCb credible.
⎨
⎩ Vcon = Zfn Icon + VCc (7) According to the derived transitive relation, the following conclu-
sions can be obtained:
where ZSn is the system impedance; Zfn is the equivalent impedance of

720
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Table 1 converged on the same horizontal line. However, being converged on a

The limit of the matrix elements in TI and TV. constant means that the right parts (phasor sum) of (9) and (10) become
Matrices Matrix Operating Mode zero, and the grid-side harmonic currents will be in a balanced state,
Elements that is, few harmonic currents can be transferred into the grid winding.
No Filter IPF Mode CIPF Mode According to the above analysis, the mathematical interpretation of the
(|Zfn|→∞) (Kn → 0) (Kn→∞) CIPF system for harmonic suppression can be revealed as follows: when
TI Δ1 N2 N3 ZP1 ZP2 − ZP1 implementing the CIPF method, with the increase of Kn, the first ele-
Δ -( + ) −
N1 3N1 ZA 3ZA ment in matrix TI or TV (i.e., |Δ1/Δ| or |Δ4/Δ|) which represents the
Δ2 N3 ZP2 ZP2 − ZP1
Δ
3N1 ZA 3ZA
transfer proportion of the identical phase harmonic will reduce to a
Δ3 0 0 ZP2 − ZP1 non-zero value. To achieve the balanced state, the other two elements
Δ
3ZA
(i.e., |Δ2/Δ| and |Δ3/Δ| or |Δ5/Δ| and |Δ6/Δ|) are required to be ad-
TV Δ4
Δ
0 1 1 justed (increase or reduce) to compensate the remaining part of the first
ZA 3ZA
Δ5 0 0 1 element which cannot be totally offset. Only when the three matrix
Δ
3ZA elements tend to the same, the grid-side harmonic currents can be
Δ6
Δ
0 0 1
suppressed completely. In consequence, the distinctive filtering prin-
3ZA
ciple of harmonic balance is the main difference between the proposed
CIPF and the HAPF in [21], ascribed to the special winding config-
(1) According to the preconditions I&II of dual zero impedance design uration of the new transformer.
for implementing the IPF method [26] and (A. 1), i.e., I,
Z3n + 3Zfn = 0; II, |Zfn| = 0, we can obtain that |ZP1| = |ZP2| = 0, 4.2. Establishment of evaluation criterion
ideally. That is to say, TI = 0. Hence, there is few harmonic current
transferring into the grid winding, which means that the load side Note from (9) and (10) that ISn is determined by ILn and VSn si-
harmonic currents are offset with the induced harmonic current at multaneously. Moreover, the grid-side current transferred from the two
the secondary winding. main harmonic sources is highly associated with the elements (Δ1/Δ)-
(2) The transfer matrix TI in (9) is different in form with the one TL in (Δ6/Δ). To explore the effect of the virtual impedance on the filtering
(3). When the CIPF is in normal operation, for each phase, the grid- performance and provide a theoretical guidance for parameter setting,
side current is coupled with the load currents of three phases. it is of vital importance to establish an evaluation criterion which uses
While, when the IPF is in operation (i.e., Kn = 0), the grid-side the filtering rate and damping rate as the evaluation indexes. Separately
current per phase is only associated with any two load currents. considering the harmonics from the rectifier load, and shifting the
(3) It can be found from (9) and (10) that, the grid-side current is phase information of the currents into the matrix elements, (9) can be
highly associated with the matrix elements Δ1∼Δ6 and Δ. More- rewritten as
over, Δ1∼Δ6 and Δ all include Kn, which means that ISn can be
Δ 1 Δ2 2 Δ3
controlled by adjusting the virtual impedance. ⎧ ISAn = ( Δ + Δ
a + Δ
a ) ILAn
⎪ Δ Δ2 2 Δ3
ISBn = ( Δ1 + a + a ) ILBn
⎡ ISAn ⎤ ⎨ Δ Δ
⎢ ISBn ⎥ ⎪ ISCn = ( Δ1 + Δ2 2
a +
Δ3
a ) ILCn
⎢ ISCn ⎥ ⎩ Δ Δ Δ (11)
⎣ ⎦
n∙120°
2
ZA2 ZP2−ZA12
ZP1 Z2 ZP2−ZA1 ZA2 ZP1 ZP2 ZA1 ZA2−ZA2 2
ZP1 ⎤ where a = e , and n = f/f1, (f1 = 50 Hz), i.e., the 120° difference
⎡    A1
 
⎢ ⎥ between each phase.
Δ1 Δ2 Δ3
1⎢ 2 2 2 2
⎥ Separately considering the harmonics from the upstream, (10) can
= ⎢ ZP2 ZA1 ZA2−ZA2 ZP1 ZA2 ZP2−ZA1 ZP1 ZA1 ZP2 ⎥ further be rewritten as
Δ⎢ ⎥
⎢ − Z Z Z
A1 A2 P1 ⎥ Δ Δ5 2 Δ6
4
⎢ Z2 Z −Z Z Z 2
ZA2 2 ZP 2−ZA1
2
ZP1 ⎥ ⎧ ISAn = ( Δ + a + a ) VSAn
⎣ A1 P2 A1 A2 P1 ZP2 ZA1 ZA2−ZA2 ZP1 ⎦ ⎪
Δ Δ
Δ4 Δ5 2 Δ6
ISBn = ( Δ + a + a ) VSBn
⎡ ILAn ⎤ ⎡ ILAn ⎤ ⎨ Δ Δ
⎢ ILBn ⎥ = TI ⎢ ILBn ⎥ ⎪ ISCn = ( Δ4 + Δ5 2
a +
Δ6
a ) VSCn
⎢ ILCn ⎥ ⎢ ILCn ⎥ ⎩ Δ Δ Δ (12)
⎣ ⎦ ⎣ ⎦ (9)
According to (11) and (12), the evaluation indexes are listed in (13).
2
⎡ ZA1 ZA1
 ZA2
2
ZA2 ⎤ The comprehensive filtering rate (CFR, γ) or the comprehensive
⎡ ISAn ⎤ 1 ⏟
⎢ Δ4 Δ5 ⏟
Δ6 ⎥ ⎡VSAn ⎤ ⎡VSAn ⎤ damping rate (CDR, λ) uses the condition of no filter or passive mode as
⎢ ISBn ⎥ = ⎢ 2 2
⎥ ⎢ VSBn ⎥ = TV ⎢ VSBn ⎥
the baseline, respectively. A high value of CFR or CDR represents the
⎢ ISCn ⎥ Δ ⎢ ZA2 ZA1 ZA1 ZA2 ⎥ ⎢
V ⎥ ⎢VSCn ⎥
⎣ ⎦ ⎢Z Z Z 2
Z 2 ⎥ ⎣ SCn ⎦ ⎣ ⎦ satisfactory filtering performance.
⎣ A1 A2 A2 A1 ⎦ (10)
Δ Δ
1 + 2 a2 + 3 a Δ
⎧ Δ Δ Δ
⎪ γ = (1− |d1 + d2 a2| ) × 100%
⎪
4. Analysis of filtering characteristic Δ4 Δ Δ
⎨ + 5 a2 + 6 a
Δ Δ Δ
⎪ λ = (1− Δ4 Δ Δ
) × 100%
4.1. Filtering mechanism ⎪ + 5 a2 + 6 a
⎩ Δ Δ Δ Kn = 0 (13)

To comprehensively analyze the filtering principle of the CIPF

system for three-phase harmonics from the load side and the grid side,
the curves of magnitude and phase characteristic for each element in
matrices TI and TV are plotted, as shown in Fig. 5. An assumption is
made here: the three-phase harmonic currents/voltages all have the
same amplitude, and there is a difference of 120° between each phase.
It can be observed from the figures that, when Kn approaches to the
positive infinity, the three curves of magnitude and phase are all
4.3. Filtering performance analysis
Combined with the design parameters of the prototyping system
listed in Tables 2 and 4, the curves of CFR and CDR varied with the
virtual impedance are given out in Fig. 6. Noted that the time delay
caused by the controller is not taken into consideration for simulating
the ideal case. As for the general trend, the filtering system can better

721
Q. Liu et al.
Fig. 5. Magnitude and phase characteristics of the harmonic transfer matrix elements.
suppress the harmonics from the load side and grid side with the in-
crease of virtual impedance. But unfortunately, it can be further found
from the enlarged view that, the performance gets poor when Kn is quite
small. Especially for 11th harmonic, its CDR curve even drops below 0-
axis. The part below 0-axis indicates that the harmonic currents gen-
erated by the background harmonic voltage are amplified at grid side.
This is extremely unfavorable for harmonic elimination, and it is pos-
sible to excite harmonic resonance. The problem will only get even
worse if time delay is considered [28]. Moreover, it should be noted
that, a larger virtual impedance to avoid the potential amplification
area may result in system instability [29]. In conclusion, the classic
722
Electrical Power and Energy Systems 105 (2019) 717–728
resistive control method mentioned in (4) is no longer suitable for the
proposed filtering system.
5. Design of improved control strategy
5.1. Control strategy
To obtain a satisfactory harmonic elimination performance, the
curve for CFR or CDR should be monotone increasing. An improved
control strategy is proposed to cope with this issue, as shown in Fig. 7.
The control algorithm includes 5th, 7th, 11th and 13th harmonic
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Fig. 6. The curves of (a) CFR and (b) CDR without considering time delay.

detection algorithms and the corresponding virtual impedance control.

Unlike the conventional resistive control, the module of phase shift is
used to make the virtual impedance pure inductive (L), resistive-in-
ductive (R-L), pure resistive (R), resistive-capacitive (R-C) and capaci-
tive (C), so that the virtual impedance can operate in the four quadrants
according to the specific operational states. The phase shift can be
realized by the matrix T with the adjustable parameter θn in (14).
Taking the inductive control as an example, the output voltage of the
inverter is controlled to be 90° ahead of the detected harmonic current.
⎡i¯′dn ⎤ = ⎡ cos θn − sin θn ⎤ ·⎡i¯dn ⎤
⎢i¯′qn ⎥ ⎢ sin θn cos θn ⎥ ⎢i¯qn ⎥
⎣ ⎦ ⎣  ⎦ ⎣ ⎦
T (14)
Fig. 8. The relation between the evaluation indexes with module and phase of
virtual impedance. (a) For CFR. (b) For CDR.
5.2. Performance exploration for four-quadrant impedance control
By adopting the four-quadrant impedance control, this paper makes
further exploration to optimize the filtering performance. Taking 11th
harmonics as an example, Fig. 8 shows the relation between the eva-
luation indexes with the module and phase of the virtual impedance. A
time delay of interrupt cycle exists in the practical system which uses

Fig. 7. Control strategy of CIPF.

723
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

DSP as the controller, so the inherent time delay of 0.1 ms is considered Table 2
in the plot. As shown in Fig. 8, the curved surfaces of CFR and CDR are CFR or CDR with different virtual impedance control mode.
similar. A hollow exists in the module interval of [10, 30] and in the Mode Item 5th 7th 11th 13th
phase-lagged region. Around the hollow, CFR or CDR drops infinitely,
which means that the harmonic might be seriously amplified if the filter Improved Control Optimal Phase 60° 53° 130° 76°
CFR (max) 96.08% 64.19% 43.03% 38.17%
operates in this control region. Thus, this control region should be
Resistive Control CFR 95.57% 60.46% −7.41% 26.11%
avoided in actual operation. Conversely, in the phase-lead region, CFR
or CDR rises with the increase of the module, thus an excellent filtering Improved Control Optimal Phase 60° 52° 130° 76°
CDR (max) 71.60% 51.55% 32.58% 27.89%
performance can be guaranteed. Resistive Control CDR 67.96% 45.86% −29.22% 13.34%
According the aforementioned control law, Fig. 9 further gives the
evaluation curves with four-quadrant variable virtual impedance. Sev-
eral details are taken into account: (1) based on 6-pulse rectifier, the brought by the low pass filter (LPF) in the control algorithm, the actual
curves for 5th, 7th, 11th and 13th are plotted; (2) only the time delay of curve for resistive control is possible in the yellow region of Fig. 9,
0.1 ms caused by sampling and computing is considered; and (3) con- which makes the grid-side current waveform distorted.
sidering the stability margin, the module for each order harmonic is set
to be 10.
6. Experimental verification
As shown in Fig. 9, four operating regions are divided as follows: I.
L ∼ R [0°, 90°), II. L ∼ -R [90°, 180°), III. C ∼ -R [180°, 270°) and IV.
To verify the proposed CIPF method with the improved control
C ∼ R [270°, 360°). It can be observed from Fig. 9 that the filtering
strategy, a 5-kV A three-phase CIPF laboratory prototype is built.
performance can be effectively improved by adjusting the parameter θn
Fig. 10 shows the real photo of the prototyping system and its wiring
only if it is in region I. Conversely, the performance gets worse when
diagram. The parameters of the new transformer are listed in the
operating in the capacitive regions (region III or IV). Table 2 lists the
Table 3. The experimental system uses the diode-bridge rectifier and a
optimal phase corresponding to the maximum of CFR or CDR in the four
40 Ω resistor to simulate the rectifier load. DSP is adopted as the core
operating regions. It can be found that, by means of the improved
controller of the inverter, which is programmed in the environment of
control strategy, CFR and CDR rise obviously compared with the con-
CCS 3.0. The currents/voltages at the grid side and at the DC side of the
ventional resistive control, especially for the higher order harmonics. It
inverter are sampled as the input signal of the controller. HIOKI-PW
should be mentioned here that, when considering the phase delay
3198 power quality analyzer is the main measure to record the ex-
perimental data, such as the current waveform, the harmonic content
and the total harmonic distortion (THD). Agilent DSO-X 3214A oscil-
loscope is used to monitor the DC voltage of the inverter. More related
prototype parameters are given in Table 4.

6.1. Experimental 1: Filtering performance test

Fig. 11 shows the test results of filtering performance, and Table 5

lists the harmonic content of grid-side currents/voltages under steady
state. It can be found from the table that the harmonic content of the
grid voltage is quite high, which certainly makes the grid-side current
waveform distorted. The negative factors of (1) the only 5th single
tuned filter and (2) the non-negligible line impedance in such down-
scaled system together make it difficult to eliminate the harmonics. As a
consequence, when compared Fig. 11(a) with (b), the performance of
IPF mode is not as well as the results obtained in [26]. The fundamental
capacitive current is the main component existed in the filtering branch
current [see in Fig. 11(c)]. However, when the CIPF with resistive
control puts into operation, the power quality at the grid side is im-
proved and THDi drops to 12.17%. Fig. 11(d) shows a good dynamic
performance of the CIPF when load changes. The controller responds
almost simultaneously to track the load change, and the response time
is less than half a cycle.
Furthermore, as shown in Fig. 11(e), the DC capacitor of the in-
verter is not precharged, and its initial voltage is zero. After a short
period of voltage regulation, the DC voltage raises to its set value
(40 V).

6.2. Experimental 2: Four-quadrant impedance control test

As shown in Fig. 12, the experiment of the four-quadrant impedance

Fig. 9. The evaluation curves with four-quadrant variable virtual impedance. control presented in Section 5.2 is conducted, as well. According to the
(a) For CFR. (b) For CDR. data in Table 5, the filtering rates of the resistive control (θn = 0°) for
5th, 7th, 11th and 13th harmonics are 82.80%, 68.09%, 23.08% and

724
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Table 4
Prototyping system parameters.
Item Parameter Value

Power Grid Grid Voltage 409 V

Simulated System Impedance (LS) 3 mH
Power Frequency 50 Hz

Current Sensor Transformation Ratio 100 A/50 mA

Nonlinear Load Resistor (RL) 40 Ω

Passive Filter Filtering Capacitor (Cf) 248 μF

Filtering Reactor (Lf) 0.834 mH
Snubber Resistor (Rf) 10 Ω

Inverter Intelligent Power Module 1200 V/75 A

(PM50RLA120)

Controller Digital Chip TMS320F28335

Sample Frequency 10 kHz
Dead-time 2.4 μs

DC-link for Inverter Reference Voltage 40 V

DC Capacitor Bank (Cdc) 1500 μF
Non-inductive Capacitor (Cni) 1.0 μF
941C6W1K-F)

Control Algorithm Cut-off Frequency for LPF (fc) 1.0 Hz

PI Regulation for DC voltage Control P = 1.5
I = 5 × 10−5
5th Virtual Impedance (Optimal control) 10 Ω∠60°
7th Virtual Impedance (Optimal control) 10 Ω∠50°
11th Virtual Impedance (Optimal 10 Ω∠130°
control)
13th Virtual Impedance (Optimal 10 Ω∠75°
control)

28.00%. It can be further observed from Fig. 12(b) that the power
quality at the grid side is significantly improved after implementing the
optimal control, while the filtering rates increase to 92.47%, 85.11%,
66.67% and 44.00%, respectively. Moreover, as shown in Fig. 12(c),
when the filtering system operates in control region II, the filtering
performance gets even worse compared with the resistive control and
the optimal control. However, after the operating mode is switched to
the control region III, the harmonics are greatly amplified at the grid
side. The experimental results are consistent with the theoretical ana-
lysis. By means of the improved control strategy, the filter can obtain
the optimal performance.

Fig. 10. Experimental system. (a) Experimental prototype. (b) Wiring diagram. 7. Conclusion

This paper proposes the CIPF system for harmonic bi-directional

elimination nearby the non-linear load. The LC-coupling active filter is
Table 3 connected to the main transformer through the customized filtering
Design parameter of the new transformer. winding. This structure not only provides isolation and voltage
Items Grid Winding Load Winding Filtering Winding
matching, but suppresses the harmonics between the secondary wind-
ings to attenuate the negative impact of the harmful components on the
Capacity [kV A] 30 30 20 transformer. The mathematical model and the revelation of the filtering
Wiring Mode Star Extended-delta mechanism are given in detail, based on the deduced harmonic transfer
Rated Voltage [V] 230.94 100.00 100.00
matrices. According to the special operational law, an improved control
Number of Turns 104 45 45
Equivalent Impedance [%] 5.50 3.85 0.04 strategy is designed to avoid the potential harmonic enlargement. The
feasibility and effectiveness of the proposed CIPF system with its

725
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Fig. 11. Experimental waveforms of the filtering performance test. (a) Grid-side current in no filter mode. (b) Grid-side current in IPF mode. (c) Filtering branch
current when passive filter puts into operation. (d) Grid-side current in CIPF mode during load change. (e) DC-link voltage when inverter is enabled. (f) Steady-state
waveform of the DC-link voltage.

726
Q. Liu et al. Electrical Power and Energy Systems 105 (2019) 717–728

Table 5
Harmonic content under various operating mode.
Harmonic Order Grid Voltage Operating Mode

No Filter IPF Mode CIPF Mode

Resistive Control Control Region II Optimal Control

(10 Ω∠0°) (10 Ω∠135°)

Fun. 237.31 V 4.46 A 4.38 A 4.41A 4.43A 4.45A

5 4.71 V 0.93 A 0.56 A 0.16A 0.18A 0.07A
7 2.09 V 0.47 A 0.36 A 0.15A 0.19A 0.07A
11 1.43 V 0.39 A 0.31 A 0.30A 0.15A 0.13A
13 0.20 V 0.25 A 0.23 A 0.18A 0.17A 0.14A
THD 2.82% 27.00% 20.75% 12.17% 13.37% 4.11%

Fig. 12. Experimental waveforms of the four-quadrant impedance control test. (a) Grid-side current in CIPF mode with resistive control. (b) Grid-side current in CIPF
mode with optimal control. (c) Grid-side current in control regions II and III.

control strategy are verified by the laboratory prototype, in which it system is a cost-effective solution for harmonic suppression in the high-
shows a good filtering performance. In conclusion, the proposed CIPF power applications.

Appendix A

The characteristic impedances of the CIPF are as follows

N N
1 2 N1
⎧ ZP1 = N 2 (Z3n + 3Zfn ) + N3 Zfn
3
⎪
⎪ ZP2 = N1 Zfn
⎪ N3
⎪ N N1
ZA1 = ( N1 )2 (Z3n + 3Zfn ) + Z1n + ZSn + K
N3 n
⎨ 3
⎪ Z = N1 K
A2 N3 n
⎪
⎪ N1
⎪ ZA = ( N3 )2 (Z3n + 3Zfn ) + Z1n + ZSn (A.1)
⎩

References
[1] Kurohane K, Senjyu T, Yona A, Urasaki N, Goya T, Funabashi T. A hybrid smart AC/
DC power system. IEEE Trans Smart Grid 2010;1(2):199–204.
[2] Tian H, Li Y, Wang P. Hybrid AC/DC system harmonics control through grid in-
terfacing converters with low switching frequency. IEEE Trans Ind Electron
2018;65(3):2256–67.
[3] Bubshait AS, Mortezaei A, Simões MG, Busarello TDC. Power quality enhancement
for a grid connected wind turbine energy system. IEEE Trans Ind Appl
2017;53(3):2495–505.
[4] Liang X. Emerging power quality challenges due to integration of renewable energy
sources. IEEE Trans Ind Appl 2017;53(2):855–66.
[5] Chen J, Zhang X, Wen C. Harmonics attenuation and power factor correction of a

727
Q. Liu et al.
more electric aircraft power grid using active power filter. IEEE Trans Ind Electron
2016;63(12):7310–9.
[6] Mohammed OA, Abed NY, Liu S. Investigation of the harmonic behavior of three
phase transformer under nonsinusoidal operation using finite element and wavelet
packets. IEEE Trans Magn 2006;42(4):967–70.
[7] IEEE Recommended Practice for Establishing Transformer Capability When
Supplying Nonsinusoidal Load Currents, IEEE Stand C57.110-1998; 1998.
[8] Wagner VE, et al. Effects of harmonics on equipment. IEEE Trans Power Del
1993;8(2):672–80.
[9] Das JC. Passive filters—potentialities and limitations. IEEE Trans Ind Appl
2004;40(1):232–41.
[10] Orillaza JRC, Wood AR. Harmonic state-space model of a controlled TCR. IEEE
Trans Power Del 2013;28(1):197–205.
[11] Hamadi A, Rahmani S, Al-Haddad K. A hybrid passive filter configuration for VAR
control and harmonic compensation. IEEE Trans Ind Electron 2010;57(7):2419–34.
[12] Ghahremani E, Kamwa I. Optimal placement of multiple-type FACTS devices to
maximize power system loadability using a generic graphical user interface. IEEE
Trans Power Syst 2013;28(2):764–78.
[13] Hu H, He Z, Gao S. Passive filter design for China high-speed railway with con-
sidering harmonic resonance and characteristic harmonics. IEEE Trans. Power Del.
2015;30(1):505–14.
[14] Akagi H. Active harmonic filter. Proc IEEE 2005;93(12):2128–41.
[15] Jintakosonwit P, Fujita H, Akagi H. Control and performance of a fully-digital-
controlled shunt active filter for installation on a power distribution system. IEEE
Trans Power Electron 2002;17(1):132–40.
[16] Wang L, Chang C, Kuan B, Prokhorov AV. Stability improvement of a two-area
power system connected with an integrated onshore and offshore wind farm using a
STATCOM. IEEE Trans Ind Appl 2017;53(2):867–77.
[17] Sano K, Takasaki M. A transformerless D-STATCOM based on a multivoltage cas-
cade converter requiring no DC sources. IEEE Trans Power Electron
2012;27(6):2783–95.
[18] Zhang Z, Xie B, Hu S, Li Y, Luo L, Rehtanz C, et al. Reactive power compensation
and negative-sequence current suppression system for electrical railways with
YNvd-connected balance transformer—Part I: theoretical analysis. IEEE Trans
728
Electrical Power and Energy Systems 105 (2019) 717–728
Power Electron 2018;33(1):272–82.
[19] Gonzatti RB, Ferreira SC, Silva CH, Pereira RR, Silva LEB, Torres GL. Smart im-
pedance: a new way to look at hybrid filters. IEEE Trans Smart Grid
2016;7(2):837–46.
[20] Luo A, Xu X, Fang L, Fang H, Wu J, Wu C. Feedback-feedforward PI-type iterative
learning control strategy for hybrid active power filter with injection circuit. IEEE
Trans Ind Electron 2010;57(11):3767–79.
[21] Akagi H, Isozaki K. A hybrid active filter for a three-phase 12-pulse diode rectifier
used as the front end of a medium-voltage motor drive. IEEE Trans Power Electron
2012;27(1):69–77.
[22] Rahmani S, Hamadi A, Al-Haddad K, Dessaint LA. A combination of shunt hybrid
power filter and thyristor-controlled reactor for power quality. IEEE Trans Ind
Electron 2014;61(5):2152–64.
[23] Lam C, Wang L, Ho S, Wong M. Adaptive thyristor-controlled LC-hybrid active
power filter for reactive power and current harmonics compensation with switching
loss reduction. IEEE Trans Power Electron 2017;32(10):7577–90.
[24] Lei E, Yin X, Zhang Z, Chen Y. An improved transformer winding tap injection
DSTATCOM topology for medium-voltage reactive power compensation. IEEE Trans
Power Electron 2018;33(3):2113–26.
[25] Li Y, Liu Q, Hu S, Liu F, Cao Y, Luo L, et al. A virtual impedance comprehensive
control strategy for the controllably inductive power filtering system. IEEE Trans
Power Electron 2017;32(2):920–6.
[26] Li Y, Luo L, Rehtanz C, Ruberg S, Yang D, Xu J. An industrial DC power supply
system based on an inductive filtering method. IEEE Trans Ind Electron
2012;59(2):714–22.
[27] Li Y, Peng Y, Liu F, Sidorov D, Panasetsky D, Liang C, et al. A controllably inductive
filtering method with transformer-integrated linear reactor for power quality im-
provement of shipboard power system. IEEE Trans Power Del 2017;32(4):1817–27.
[28] Zhou Z, Liu Y. Time delay compensation-based fast current controller for active
power filters. IET Power Electron 2012;5(7):1164–74.
[29] Srianthumrong S, Fujita H, Akagi H. Stability analysis of a series active filter in-
tegrated with a double-series diode rectifier. IEEE Trans Power Electron
2002;17(1):117–212.