AMSE JOURNALS-AMSE IIETA publication-2017-Series: Modelling A; Vol.

90; N1; pp 98-116

Submitted Jan. 2017; Revised March 15, 2017, Accepted April 15, 2017

Model Predictive PWM for Hybrid Active Power Filter

*Yudong Li, **Shaotong Du, ***Tianyu Zhang, **** Xingguo Tan

* School of Electrical Engineering and Automation, Henan Polytechnic University,

Jiaozuo 454003, Henan Province, China, ([email protected])
** School of Electrical Engineering and Automation, Henan Polytechnic University,
Jiaozuo 454003, Henan Province, China, ([email protected])
*** School of Mechanical and Electrical Engineering, Henan Vocational College of Industry and
Information Technology, Jiaozuo 454003, Henan Province, China, ([email protected])
**** School of Electrical Engineering and Automation, Henan Polytechnic University,
Jiaozuo 454003, Henan Province, China, ([email protected])

Abstract
Model predictive pulse-width modulation (MP-PWM) is applied to hybrid active power filter
(HAPF) in this paper. The configuration, control, carried-based PWM (CPWM) and filtering
performance of HAPF are described in this paper. To realize the MP-PWM, the discrete-time
model of the inverter is presented and a cost function is built. The flow chart of realizing the MP-
PWM is presented. The PWM is assumed as a sampling-hold circuit with an equivalent control
period to compare MP-PWM and CPWM. To compare the equivalent control period of PWM, an
inverter simulation with R-L-C load is carried out. The inverter simulation results verify that the
equivalent control period of MP-PWM is smaller than the CPWM, while the average switching
frequency is lower than the CPWM. The simulation results of HAPF with MP-PWM and CPWM
show that the filter performance of HAPF with the MP-PWM is better. An experiment study of
HAPF with MP-PWM has been performed to validate the MP-PWM for HAPF.

Key words
Harmonic suppression, Hybrid active power filter, Model predictive, Pulse width modulation
(PWM)

98
1. Introduction
With the large input of nonlinear loads, the harmonic problem of the power system is
becoming increasingly serious. Hybrid active power filter ( HAPF), characterized with smaller
rating of active power filter (APF), is one of the effective ways for harmonic suppression, which
is easier to eliminate harmonics of the high-voltage grid [1]-[3].
The typical HAPF topology contains a passive power filter (PPF) and an APF, the two of
which can be connected either directly or through a coupling transformer [3]. Many types of
HAPF topology have been reported already [4]-[9]. Injecting the APFs current into the PPF
which contains a capacitor and a fundamental resonant LC filter through a coupling transformer
can greatly reduce the rating of the APF [4]-[5], But this type HAPF needs a dc power to supply
the dc capacitor of the inverter. Two H -bridges are connected in series with the grid via coupling
transformers which have different voltage ratios to achieve the nine-level HAPF, which can
obtain high-quality voltage waveforms and reduce the switching loss [6]. The HAPF, an inverter
connected with a single-tuned LC filter in series, can realize the harmonic current suppression of
the diode rectifier more economically [7]-[8], which is a good scheme for the grid whose voltage
is lower than 6kV. A parallel HAPF was proposed in [9], in which two inverters have different
switching frequencies.
The control methods of HAPF are attracting scholars attention. Proportional resonant
PR controller was applied to the HAPF to achieve frequency-selective harmonic suppression
[10]. Feed forward control was applied to the HAPF to improve the systems compensation
performance [11], but only single order harmonic filter performance is improved in this way.
Moreover, sliding mode control [12] and fuzzy control [13]-[14] were also applied to the HAPF.
A double closed-current control was proposed to reduce the fundamental current of the inverter in
the injection-type HAPF [15]. Hysteresis current control in [16] was applied to a three-phase
four-wire HAPF.
In recent years, the model predictive control (MPC) has been widely used in various
converters [17].The cost function and the discrete-time model are the key components of the
MPC. The optimizing is carried out based on the value of cost function. The MPC has been
applied to current control, power control, torque and flux control, capacitor voltage balancing
control, etc [18]. The current control using MPC is widely used in the rectifier [19]-[20], the
active power filter [21]-[22], motor drives [23], etc. The predictive control can increase the
bandwidth of the active power filter [24]. The MPC has also been tried on HAPF [25], but the
HAPF with more reactors and capacitors makes the model complex. The selective harmonic

99
elimination PWM using MPC was mentioned in [26], but the weighting factor design is difficult
to be realized in HAPF, because the reference voltage of inverter is not sinusoidal. In this paper,
the topology structure reported in [7] is adopted to suppress the harmonic pollution caused by the
diode rectifier loads. The model predictive pulse-width modulation (MP-PWM) which is easy to
be digitally realized and only needs discrete-time model of the converter is presented [27]. The
HAPF with MP-PWM is studied. The MP-PWM for HAPF can have high sampling frequency.
Meanwhile, the average switching frequency is lower than the sampling frequency.
The configuration control and carrier-based PWM (CPWM) [28] of the HAPF are
described in this paper. Meanwhile, the difference of filtering performance caused by CPWM
using symmetrical regular sampling and asymmetric regular sampling is analyzed. The MP-PWM
which contains discrete-time model, cost function and voltage optimizing is presented in detail.
The MP-PWM is compared with CPWM by simulation. To provide the validity of the MP-PWM
for HAPF in this paper, experiment results obtained with the diode rectifier load are provided.

2. System Description
2.1 System Configuration
Fig.1 shows the configuration of the HAPF considered in this paper. The HAPF topology
consists of a LC filter which is almost resonant at 5th harmonic frequency and a converter. The
HAPF is connected at the point of common connection (PCC). In Fig.1, LS represents the supply
inductance; iSx, iFx and iLx (x=A, B, C)are the supply current , compensation current and load
current respectively; eSx is the supply voltage; ex is the phase voltage at the PCC; C and L
represent the capacitor and the inductance of the LC filter, respectively; Sy stands for the power
switch device of the converter (y=1, 2, 3, 4, 5, 6) ; Cdc is dc-side capacitor and Udc is the dc-bus
voltage.

eSA LS iSA eA iLA

eSB iSB eB iLB nonlinear
eSC iSC eC iLC loads
o
50Hz iFA iFB iFC eA
PLL U*dc +
C
- PI
sin(1t) and cos(1t) Udc
L generators
S1 S3 S5 iSA iSB iSC
C +
Udc +
A
B
Cdc - iSA
abc-dq iq i-q - dq-abc iSA1- iSAH u*A
LPF
LC filter iSB transform -
+ Trans- iSB1 -+ iSBH K u*B P 6
S2 S4 S6 id id form iSC1 + W
N iSC LPF - iSCH u*C M
converter
hybrid active power filter

100
 Fig. 1. Configuration of the HAPF Fig.2. Diagram of the HAPF control

2.2 System Control and Carrier-Based PWM

The control system of the HAPF is shown in Fig. 2. The harmonic detection method based
on the instantaneous reactive power theory is utilized to detect the harmonic current of the supply
current [7]. The abc-dq transformation converts the supply currents iSA, iSB and iSC into the
instantaneous active current id and the instantaneous reactive current iq. The fundamental
components in the supply currents correspond to dc components in id and iq. Two fourth order
Butterworth low-pass filters with the same cutoff frequency of 30Hz extract dc

iq id
components and from id and iq, respectively. The dq - abc transformation produces the supply
fundamental currents iSA1, iSB1 and iSC1. Then, subtract the supply fundamental currents iSA1, iSB1
and iSC1 from the supply currents iSA, iSB and iSC to gain the supply harmonic currents iSAH, iSBH
and iSCH, respectively. Multiply the supply harmonic current iSxH by the gain of K to get the
reference phase voltage of the converter KiSxH. The fundamental current through the HAPF is

leading, so injecting the iq into the dc components iq can regulate the dc capacitor voltage of the
converter [7]. A proportional integral (PI) control is used as the control algorithm. Finally, the
PWM gate signals for IGBTs can be calculated with the PWM algorithm.
The carrier-based PWM (CPWM) is widely used [28].When the digital signal processor is
used to realize carrier-based PWM, symmetrical regular sampling and asymmetric regular
sampling are widely used [30], which are shown in Fig. 3(a) and Fig. 3(b), respectively. Ts is the
sampling period in Fig.3. In the case of symmetrical regular sampling, Fig. 3(a), sampling
frequency of reference voltage is equal to the frequency. From Fig. 3(b), asymmetric regular
sampling, sampling frequency is twice of the carrier frequency.

(k-1)TS kTS (k+1)TS t (k-2)TS kTS (k+2)TS t

(a) (b)
Fig.3. Sampling techniques. (a) Symmetrical regular sampling. (b) Asymmetric regular sampling

101
3. Filtering Characteristic Analysis
The reference voltage of the converter is KiSxH in harmonic frequencies. Neglecting the
harmonics produced by on-off power device, the PWM converter can be assumed as a sampling
and hold circuit with a transfer H(s)

1 1 e sTC
H s (1)
TC s

where TC is equivalent control period and s is the Laplace operator, TC is related to delay time.

iSxH ISxH(s)
iFxH IFxH(s)
1
LS C sLS ILxH(s)
sC
iLxH
L sL
+ +
uxH UxH(s)
- -
(a) (b)
Fig. 4. Diagram of the single-phase circuit of the HAPF in the harmonic
domain.(a)equivalent circuit.(b) equivalent Laplace operation circuit

Table 1. Parameters of the HAPF

LS 0.1mH
L 4mH
C 100F
K 30

To analyze the filtering characteristic of HAPF, a single- phase equivalent circuit of the
HAPF in the harmonic domain is shown in Fig. 4(a). The uxH is the inverters output voltage. The
Laplace operation circuit of the Fig. 4(a) is shown as Fig. 4(b). In Fig. 4(b), ISxH(s), IFxH(s) and
ILxH(s) indicate the image functions of the grid harmonic current iSxH, the compensation harmonic
current iFxH and the load harmonic current iLxH, respectively; UxH(s) is the image function of uxH.

U xH s H s KI SxH s (2)

102
According to Kirchholffs laws, the circuit equations from Fig. 4(b) can be obtained as

IS xH s I F xH s I L xH s

1
sLS IS xH s sL I F xH s U xH s 0
(3)
sC

From (2) and (3), the transfer function G(s) is given as

ISxH s LCs 2 1
(4)
I LxH s LS L Cs 2 KCs H s 1
The characteristic of the HAPF can be observed by the bode plot of (4). The parameters shown in
table 1 are adopted in this paper. From (4) and table 1, the bode plots are shown as Fig. 5.
From Fig. 5, the resonance peak is reduced when the equivalent control period TC of PWM is
smaller. In other words, harmonic amplification is reduced, the filtering performance is better. So,
short equivalent control period for PWM is helpful to improve filtering performance of the HAPF.

20
Magnitude (dB)

-20
TC=100s
TC=200s
-40
2 3
10 10
Frequency (Hz)
Fig. 5. Bode plots of the transfer function G(s) with TC=100s and TC=200s

We can know that HAPF can effectively suppress the harmonic of the supply current when K
is large enough [7]. But in consideration of the stability of the HAPF control system, the value of
K shouldnt be increased unlimitedly. The harmonic impedance of LC filter is small near the
resonant frequency, but it increases when the frequency is far away from the resonant frequency.
Therefore, when K is constant, the harmonic suppression capability of HAPF to the different
frequency harmonic current of iLx varies with the changes of LC filter harmonic impedance. The
characteristic is shown as in Fig. 5. In order to improve the harmonic suppression capability of
HAPF, the proportion K can be updated to be the proportional resonant (PR) compensator [30].
The transfer function in the frequency domain of PR is given as follows

103
 h5,7,...
2 K Ihch s
GPR ( s ) K P h s 2ch s h2
2 (5)

where KP is the proportional gain, KIh is the hth integral gain, h is the hth harmonic angular
frequency, and ch is the hth cutoff frequency. KP=30, KIh=70(h=7, 11, 13, 17, 19, 23, 25) and
h=1.5h are adopted in this paper. Replacing K with GPR(s), the bode plots are shown as Fig.6.
From Fig.6, the filtering performance is improved. And filtering performance is also becomes
better when TC of PWM is smaller.

20
Magnitude (dB)

-20 TC=100s
TC=200s
-40 3
2 10
10
Frequency (Hz)
Fig. 6. Bode plots of the transfer function with TC=100s and TC=200s when PR is adopted

4. Model predictive PWM Algorithm

The sampling techniques of Carrier-based PWM are shown in Fig. 3. The difference of
filtering performance between TC=100s and TC=200s is also analyzed. Simulation results in
part will show that equivalent control periods of the carried-based PWM using symmetrical
regular sampling (CPWM-S) and asymmetrical regular sampling (CPWM-A) are the carrier
period and twice the carrier period, respectively. To reduce control period TC of the CPWM, the
carrier frequency is increased. However, in that way the switching frequency is also increased. In
this paper, the MP-PWM producing smaller equivalent control period under the same average
switching frequency is applied to HAPF. The MP-PWM algorithm includes discrete converter
model, cost function and selecting optimal voltage vector [27].

4.1 Discrete Converter Model

The three-phase converter is shown in Fig.1. The switching state of the power switches Sy
can be represented by the switching signals sA , sB ,and sC defined as follows:

1 (if S1 on and S 2 off)

sA (6)
0 (if S1 off and S 2 on)

104
 1 (if S3 on and S 4 off)
sB (7)
0 (if S3 off and S 4 on)
1 (if S5 on and S6 off)
sC (8)
0 (if S5 off and S6 on)

From (6), (7) and (8), the converters output voltage is

uxN sxU dc (9)

where uxN is the phase-to-point (N) voltage of the converter. The converters output voltage
vector can be expressed as

2
u (uAN uBNe j 2 /3 uCNe j 4 /3 ) (10)
3

The corresponding relations between the switching signals and the converters output
voltage vector can be calculated by (6)-(10).

Table 2. Switching Signals And Voltage Vectors

(sA sB sC) 000( u1 ) 100( u2 ) 110( u3 )
voltage vector 0 U Ue
j 3

(sA sB sC) 010( u4 ) 011( u5 ) 001( u6 )

voltage vector Ue
j2 3
Ue
j
Ue
j4 3

(sA sB sC) 101( u7 ) 111( u8 )

voltage vector Ue
j5 3
0

4.2 Cost Function

* * *
The reference voltages uA , uB and uC are transformed from the abc-frame to the -frame

1 1 *
1 u A
u 2
*
2 2 *
* u B
(11)
u 3 0 3 3 u *
2 2 C

The reference voltage vector can be expressed as

u* u* ju* (12)

105
 In the MP-PWM algorithm, only one voltage vector which minimizes the cost function will
be chosen within a sampling interval. The established cost function is

g u*dt udt (13)

From (13), the discrete-time equation can be obtained as

k 1 k
g (k ) u * (m)T u(m)T
m 0
S
m 0
S (14)

where TS is the sampling period. In (14), the sampling interval TS is sufficiently small, it can
be assumed that u*(k+1)=u(k).The initial conditions in (14) are

u * (0) 0
(15)
u(0) 0

From (14) and (15), the converter output voltage vector u(k) is only uncertain in the kth
sampling interval.

4.3 Selecting Optimal Voltage Vector

The cost function (14) is as a criterion to select optimal voltage vector in the kth sampling
interval from table 2. The flow diagram of selecting optimal voltage vector is shown in Fig. 7. In
Fig. 7, the intermediate variables P(k) is

k 1 k 1
P k u *(m)TS u(m)TS (16)
m0 m0

Substituting (16) into (14), the g(k) is

g (k ) P k u k TS (17)

106
 The execution flow is as follows:
Step 1: Interrupt response, calculate variables P(k) and assign initial values to the
intermediate variables.
Step 2: With the objective to minimize the value of cost function g(k), optimization is
conducted to vectors in table 2 to search the optimal voltage vector at the kth sampling interval.
Step 3: If the zero voltage vector is chosen in step 2, redundant zero vectors are chosen,
which is based on the number of commutations. When the converter output voltage vector u(k-1)
is u1(000) , u2(100) , u4(010) or u6(001), the u1(000) should be the optimal voltage vector in the
kth sampling time interval; otherwise, the u8(111) should be the optimal voltage vector.
Step 4: Apply optimal voltage vector, the interrupt is stopped.

Startup

Calculate p(k) (16)

N
n=1
Y
Step 1 g0= , r=0
u(k-1)=u1 or N
u2 or u4 or u6
r++
Y
n=1 n=8
Substitute u(k)=ur
into (15) to calculate Step3
g(k)
Apply optimal
vector
N (u(k)=un)
g(k)<g0
Y
End
n=r; g0=g(k)

Step4
N
r>=7
Y
Step 2

Fig. 7. Flow chart of the proposed PWM

5. Simulation Results
In order to analyze the characteristics of the MP- PWM and verify its effectiveness,
simulation is carried out in this paper. Switching frequency of MP-PWM is varying .To observe
the switching frequency characteristics, the average switching frequency is thus defined as follow
[27]

107
 wa wb wc 1
f (18)
3 T

where wa, wb and wc are respectively the total number of rising edges of the signals sA, sB and sC
at the time of T, T=20ms.

5.1 Inverter Results

Matlab simulation software is adopted to simulate and analyze the results. To analyze the
equivalent control period TC of PWM, an inverter circuit simulation was set up. The inverter
circuit is shown as Fig. 8. In Fig. 8, the L and C are resonant at 550Hz. The reference voltages of
PWM are three- phase symmetrical sinusoidal voltages. The frequency of reference voltages is
550Hz. In the invert circuit, the load impedance is 1 at 550Hz, so the load current is the
expression of converters output voltage at 550Hz. The load impedance is large at other
frequencies, so the load current is almost sinusoidal.
The performances of CPWM and MP-PWM have been first verified via computer simulation,
respectively. The carrier frequency and sampling frequency 1/TS of MP-PWM are 2.5 kHz and 10
kHz, respectively. Simulation waveforms are shown in Fig. 9. In Fig. 9(a), f1 and f2 are the
average switching frequency of CPWM and MP-PWM, respectively. From Fig. 9(a), the average
switching frequency of MP-PWM f2 is lower.
The reference voltage wave and load current waves are shown in Fig. 9(b). From Fig. 9(b),
the reference voltage phase is lead the load current phase, however, the load current phase with
MP-PWM is lead the load currents phase with CPWM-A and CPWM-S.
The response curves of transfer H(s) with TC=130s and TC=200s under inputting the A-
phase reference voltage are curves q1 and q3 shown as in Fig. 9(c), respectively. The load
currents of MP-PWM and CPWM-A are curves q2 and q4, respectively. The q1 and q2 are
almost overlapping. So as q3 and q4. From Fig. 9(c), the MP-PWM is equivalent to the H(s) with
TC=130s and the S-CPWM is equivalent to the H(s) with TC=200s. Inverter simulation results
prove that the MP-PWM with lower switching frequency has smaller equivalent control period TC.

108
 R=1 L=8.374mH
C=10F Udc=200V
S1 S3 S5 L C R
+ iA
Udc iB
- iC
S2 S4 S6

Fig. 8. The simulation inverter circuit

Reference voltage(50V/div),
Load current iA(50A/div)

Load current iA(50A/div)

100
2.5 q1

Voltage(50V/div),
f1 q2
50
f(0.5kHz/div)

0 q3
1.5 q4
0
f2
Time(0.5ms/div) -50
Reference voltage
0 Load current iA under CPWM-A -100
Time(50ms/div) Load current iA under CPWM-S Time(0.5ms/div)
Load current iA under MP-PWM

(a) (b) (c)

Fig. 9. The waveforms of inverter simulation (CPWM-A and CPWM-S with carrier frequency are
2.5kHz, MP-PWM with sampling frequency is 10kHz). (a) Average switching frequency. (b)
Reference voltage and load currents. (c) Response curves of H(s) with TC=130s and TC=200s
under inputting the reference voltage and load currents of CPWM-A and MP-PWM

5.2 HAPF Results

The fully mechanized coal mining face supply voltage and the six-pulse diode rectifiers are
chosen for the simulation, where the rectifiers DC-Bus is RC load. The PR compensation is
adopted. The HAPF simulation with MP-PWM, CPWM-A and CPWM-S are carried out,
respectively. The key parameters of the simulation are shown in table 3.

Table 3. Simulation parameters

parameters value
Grid line voltage (RMS) 1140V
Supply inductance: LS 0.1mH
DC capacitance of the rectifier 3000F
Load resistance of the rectifier 10
AC inductor of the rectifier 1.5mH
Filter inductor : L 4mH
Filter capacitor: C 100F
DC voltage of the active filter: Udc 600V
DC capacitance of the converterCdc 1000F
sampling frequency of MP-PWM 10kHz
carrier frequency of CPWM-S 5kHz
carrier frequency of CPWM-A 2.5kHz

109
 Fig. 10 shows that the simulation waveforms of the HAPF with the MP-PWM. The HAPF is
switched into the grid and only regulates the dc capacitor voltage at the beginning, and then
harmonic suppression begins at 0.3s in Fig. 10. Fig. 10(a) shows the waveform diagrams of the
load three-phase currents iLA, iLB and iLC. Fig. 11(a) shows the spectrum of iLA. Fig. 10 (b) shows
the waveform diagrams of the grid three-phase currents iSA, iSB and iSC. Fig. 11(b) shows the
spectrum of iSA when the harmonic suppression is steady. Fig. 10(c) shows the waveform
diagram of the dc bus voltage Udc. Fig. 10(d) shows the waveform diagram of the average
switching frequency f with T=20ms. Fig. 11(c) and Fig. 11(d) show the spectrum diagrams of the
grid a-phase currents when the HAPF with CPWM-S and CPWM-A is switched into the grid,
respectively. The data of the special order components (1st, 5th, 7th, 11th, 13th, 17th, 19th, 23th, 25th,
29th, 31th, 35th orders) of the load a-phase current and the grid a-phase currents are listed in table 4
when the HAPF with MP-PWM, CPWM-A and CPWM-S is switched into the grid, respectively.
From Fig. 10(a) and 10(b), we can see that the grid currents become sinusoidal after the
harmonic suppression is begins and that its dynamical filtering performance is good. From
Fig.10(c), the dc bus voltage is regulated at 600V. From Fig. 10(d), the average switching
frequency is changing but its maximum value is approximately 2.5 kHz.
Comparing Fig.11(a) with 11(b), We can see the total harmonic distortion THDs of the load
current and the grid current are 27.38% and 2.73%, respectively. The filtering performance of the
HAPF with MP- PWM is good. Moreover, the harmonic components (5th, 7th, 11th, 13th, 17th, 19th,
23th, 25th orders) of grid currents are all reduced shown the table 4. From Fig. 11(c), the THD of
grid current is 2.90% when the CPWM-S with carrier frequency being 5 kHz. From Fig. 11(d),
the THD of grid current is 3.15% when the CPWM-A with carrier frequency being 2.5 kHz.
From Fig. 11 and table 4, we can see that the filtering performance of the HAPF with MP-PWM
is the best.

200 iLA iLB iLC iSA iSB iSC

200
Grid Current (A)
Load Current (A)

100 100

0 0
-100
-100
-200
-200 0.25 0.27 0.29 0.31 0.33 0.35
0.25 0.27 0.29 0.31 0.33 0.35
Time(s) Time(s)

(a) (b)

110
 800 3
2.5
600
2
Udc(V)

f(kHz)
400 1.5
1
200 0.5
0 0
0 0.1 0.2 0.3 0.4 0.5 0.2 0.3 0.4 0.5 0.6
Time(s) Time(s)

(c) (d)
Fig. 10. The simulation waveforms of the HAPF with MP-PWM. (a)
Waveforms of the load currents. (b) Waveforms of the grid currents. (c) The
waveform of dc-bus voltage. (d) The average switching frequency with T=20ms

THD= 27.38% THD= 2.90%

Mag (% of Fundamental)
Mag (% of Fundamental)

40 4
100%
100%
30 3
CPWM-S with f=5kHZ
20 2

10 1

0 0
0 10 20 30 40 0 20 40 60 80 100
Harmonic order Harmonic order
(a) (c)
THD= 2.73% THD= 3.15%
Mag (% of Fundamental)

4 4
Mag (% of Fundamental)

100% 100%
3 3
MP-PWM with f is CPWM-A with f=2.5kHZ
2 shown as Fig. 10(d) 2

1 1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Harmonic order Harmonic order
(b) (d)
Fig.11. The simulation spectrum diagrams of the load a-phase current and grid a-phase currents.
(a) The spectrum of load a-phase current. (b) The spectrum of a-phase grid current with MP-
PWM. (c) The spectrum of a-phase grid current with CPWM-S. (d) The spectrum of a-phase grid
current with CPWM- A

Table 4. Comparative Data of the load a-phase current and the grid a-phase current (peak, in
amperes)
1- 5- 7- 11- 13- 17-
order order order order order order
load current 161.45 41.27 12.17 7.40 5.16 2.39
MP-PWM 155.42 0.58 0.50 0.93 0.83 0.69
CPWM-A 155.12 1.23 0.72 0.93 1.03 0.7

111
 CPWM-S 155.65 1.08 0.7 0.89 0.99 0.69
19 23- 25 29 31 35
-order order -order -order -order -order
load current 2.20 1.50 1.18 0.95 0.84 0.63
MP-PWM 0.79 0.74 0.73 1.85 1.54 0.81
CPWM-A 0.83 0.70 0.69 2.51 1.81 1.15
CPWM-S 0.81 0.68 0.66 2.44 1.74 1.16

6. Experiment Results
In this paper, experimental verification is conducted on HAPF with MP-PWM to verify the
feasibility of this algorithm in laboratory. The power switch of the converter chosen for the
experiment is BSM50GB120DLC. The digital signal processor TMS320F2812 is employed. The
supply line voltage is 100V. The DC-bus voltage of the converter is 30V. The load is six-pulse
diode rectifier which contains an AC inductance and DC resistor. AC inductance and DC resistor
of rectifier are 3.1mH and 20, respectively. The inductance and capacitor of the PPF are 4mH
and 100F, respectively. The FLUKE435 are also employed. The sampling frequency is 10 kHz.

(a) (b) (c)

112
 (d) (e) (f)
2

f(0.5kHz/div)
1.5

0.5

0
Time(4s/div)
(g)
Fig.12. The waveforms of HAPF with MP-PWM.
(a) The waveforms of the load currents. (b) The spectrum of load a-phase current. (c) The
waveforms of the grid current when HAPF with MP-PWM is switched into the grid. (d) The
spectrum of grid a-phase current when HAPF with MP-PWM is switched into the grid.(e) The
waveforms of grid a-phase voltage and grid a-phase current when HAPF with MP-PWM is
switched into the grid.(f) The power when HAPF with MP-PWM is switched into the grid. (g)
The average switching frequency of MP-PWM

Fig. 12 describes the experimental waveforms. Fig. 12 (a) and 12(b) show respectively the
load current and spectrum of load a-phase current, which indicates serious current distortion;
Fig.12 (c) and 12(d) are respectively the grid current waveform and spectrum of grid a-phase
current when HAPF with MP-PWM is switched into the grid. It can be seen from the Fig.12(c)
that the current approaches to the sine. From Fig.12(b) and 12(d), the THD of a-phase current is
reduced to 3.2% at the grid side from 20.8% at the load side. Fig.12 (e) is the grid a-phase
voltage and the grid a-phase current. Fig.12(f) is the power and power factor. From Fig.12(e) and
12(f), the power factor is almost 1. The data of average switching frequency of MP-PWM which
is calculated in DSP is plotted as Fig.12(f) using Matlab. It can be seen from Fig.12(f) that the
average switching frequency is approximately 1.5 kHz and its maximum value approaches 1.8
kHz. Fig.12 indicates that HAPF with MP-PWM has effective harmonic compensation.

7. Conclusions
The system and filtering principle based on a kind of HAPF are analyzed in this paper. An
MP-PWM for HAPF is presented. The inverter simulation results have demonstrated the
equivalent control time of MP-PWM is smaller than CPWM-A and CPWM-S when the average
switching frequency of MP-PWM is 1.5kHz, the carrier frequency of CPWM is 2.5kHz. The
HAPF simulation results have demonstrated the filtering performance of the HAPF with MP-
PWM is better than CPWM when the carrier frequencies of CPWM-S and CPWM-A are
respectively 5kHz and 2.5kHz, the maxim average switching frequency of MP-PWM is almost

113
2.5kHz. The THD of load currents is 20.8% in the experiment. Results of the experiment show
the THD of grid currents is reduced to 3.2%, the power factor is 1 when the HAPF with MP-
PWM whose average switching frequency is approximately 1.5 kHz is switched into the grid.

Acknowledgments
This work is supported by Open Project of Key Laboratory of Control Engineering of Henan
Province (KG2016-05) and Key scientific research project of Henan higher educationGrant
No.16A470009).

References
1. P.F. Zheng, H. Akagi, and A. Nabae, A new approach to harmonic compensation in power
systems-a combined system of shunt passive and series active filters, 1990, IEEE
Transactions on Industry Applications, Vol. 26, No.6,pp. 983-990,
2. H. Akagi, Active harmonic filters, 2005, Proceedings of the IEEE, Vol.93, No.12, pp. 2128-
2141.
3. J. H. Sung, S. Park, and K. Nam, New hybrid parallel active filter configuration minimising
active filter size, 2000, IEE Proceedings - Electric Power Applications, Vol.147, No.2, pp.93-
98.
4. S. Z. Kang, L. An, Z. W. Ji et. al. Study on a novel hybrid active power filter applied to a
High-Voltage grid, 2007, IEEE Transactions on Industry Applications, Vol.54, No.5, pp.
2697-2704.
5. L. An, T. Ci, S. Z. Kang et. al. A Novel Three-Phase Hybrid Active Power Filter With a
Series Resonance Circuit Tuned at the Fundamental Frequency, 2009 , IEEE Transactions on
Industry Applications, Vol.56, No.7, pp. 2431-2440.
6. A. Varschavsky, J.Dixon, M.Rotella, and L.Morn, Cascad- -ed nine-level inverter for
hybrid-series active power filter, using industrial controller2010IEEE Transactions on
Industry ApplicationsVol.57, No.8, pp.2761-2767.
7. H.Akagi, and K.Isozaki, A hybrid active filter for a three-phase 12-pulse diode rectifier used
as the front end of a medium-voltage motor drive, 2012, IEEE Transactions on Power
Electronics, Vol. 27,No.1, pp.69-77.

114
8. N.Hatti, K. Hasegawa, and H.Akagi, A 6.6-kV Transformerless Motor Drive Using a Five-
Level Diode-Clamped PWM Inverter for Energy Savings of Pumps and Blowers, 2009, IEEE
Transactions on Power Electronics, Vol. 24,No.3, pp:796-803.
9. A. Bhattacharya, C. Chakraborty, and S. Bhattacharya, Parallel-Connected Shunt Hybrid
Active Power Filters Operating at Different Switching Frequencies for Improved Performance,
2012, IEEE Transactions on Industrial Electronics.Vol.59, No.11, pp.4007-4018.
10. L. An, P. S. Jian, W. C. Ping et. al., Power Electronic Hybrid System for Load Balancing
Compensation and Frequency-Selective Harmonic Suppression, 2012, IEEE Transactions on
Industrial Electronics, Vol.59, No.2, pp.723-732.
11. S. Srianthumrong, and H.Akagi, A medium-voltage transformerless AC/DC power
conversion system consisting of a diode rectifier and a shunt hybrid filter, 2003, IEEE
Transactions on Industry Applications. Vol.39, No.3, pp/ 983-990.
12. A. Hamadi, S. Rahmani and K. Al-Haddad, Sliding mode control of Three-Phase Shunt
Hybrid Power Filter for Current Harmonics Compensation, 2010, Proceeding of IEEE_ISIE,
pp. 1076-1082.
13. W. H. Choi, C. S. Lam, M. C.Wong, An Adaptive Source Current THD Oriented Fuzzy Logic
Controller in Hybrid Power Filter, 2011, Proceeding of IEEE_APEC, pp. 1557-1563.
14. L. An, S. Z. Kang, Z. W. Jin et. al., Development of Hybrid Active Power Filter Based on the
Adaptive Fuzzy Dividing Frequency-Control Method, 2009, IEEE Transactions on Power
Delivery, Vol. 24,No.1, pp.424-432.
15. S. Z. Kang, L. An, S. John et. al., Double Closed-Loop Control Method for Injection-Type
Hybrid Active Power Filter, 2011, IEEE Transactions on Power Electronics, Vol. 26,No.9,
pp.2393-2403.
16. C. S. Lam, M. C. Wong, Y.D. Han, Hysteresis current control of hybrid active power filters,
2012, IET Power Electronics, Vol.5, No.7, pp. 1175-1187.
17. J. Rodriguez, M. P. Kazmierkowski, J. R.Espinoza et.al., State of the Art of Finite Control Set
Model Predictive Control in Power Electronics, 2013, IEEE Transactions on Industrial
Informatics, Vol.9, No.2, pp.1003-1016.
18. J. Rodriguez, C. Patricio, Predictive control of power converters and electrical drives, 2012,
John Widely & Sons, pp.31-33
19. J. Rodriguez, J. Pontt, A. C. Silva, et al., Predictive current control of a voltage source
inverter, 2007, IEEE Transactions on Industrial Electronics, Vol. 54, No.1, pp.495-503.

115
20. R. Vargas, P. Cortes, U. Ammann et. al., Predictive Control of a Three-Phase Neutral-Point-
Clamped Inverter, 2007, IEEE Transactions on Industrial Electronics, Vol.54, No.5, pp. 2697-
2705.
21. J. D. Barros, J. F. Silva, Optimal Predictive Control of Three-Phase NPC Multilevel
Converter for Power Quality Applications, 2008, IEEE Transactions on Industrial Electronics,
Vol.55, No.10, pp.3670-3681.
22. M. Rivera, V. Yaramasu, A. Llor et. al., Digital Predictive Current Control of a Three-Phase
Four-Leg Inverter, 2013, IEEE Transactions on Industrial Electronics, Vol.60, No.11, pp.490
3-4912.
23. K. B. Lee and F. Blaabjerg, Improved sensorless vector control for induction motor drives fed
by a matri converter using nonlinear modeling and disturbance observer, 2006, IEEE
Transactions on Energy Conversions, Vol.21, No.1, pp.52 -59.
24. F. Defa
, A.-M. Llor, M. Fadel, A Predictive Control With Flying Capacitor Balancing of a
Multicell Active Power Filter, 2008, IEEE Transactions on Industrial Electronics, Vol.55,
No.9, pp.3212 -3220.
25. H. Geng, G. Yang, X. David et.al., Model Predictive Control for Transformerless Shunt
Hybrid Power Filters, 2010, Proceeding of IEEE_ IPEC , pp. 2265-2270.
26. S. Kouro, B.L.Rocca. P. Cortes et.al., Predictive control based selective harmonic elimination
with low switching frequency for multilevel converters, 2009, Proceeding of IEEE_ ECCE ,
pp. 3130-3136.
27. D. Shaotong, W. Xiaojie, Z. Juan et. al. A Novel Algorithm of PWM using virtual flux model
prediction, 2015, Proceeds of the CSEE, Vol. 35, No.3, pp. 688-694.
28. J. Holtz, Pulse width Modulation for electronic power conversion, 1994, Proceeds of the
IEEE, Vol. 82, No.8, pp. 1194-1214.
29. Lianjun Hu, Ling Tang, Quan Pan et. al., Research and Analysis of PI Control Strategy Based
on Neural Network in Power Grid, 2016, Mathematical Modelling of Engineering Problems,
Vol.3, No.1, pp.25-29.
30. R. Teodorescu, F. Blaabjerg, M. Liserre et al, Proportional- resonant controllers and filters for
grid-connected voltage-source converters, 2006, IEE Proceedings-Electric Power
Applications, Vol.153, No.5, pp.750-762.
31. Hua Li and Le L, Application of Fuzzy Control in PV- Storage Distributed Generation
System, 2016, Review of Computer Engineering Studies, Vol. 3, No. 1, pp.11-15.

116