Sunteți pe pagina 1din 10

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

Model Predictive Control and Direct Power Control


for PWM Rectifiers With Active Power Ripple
Minimization
Yongchang Zhang, Member, IEEE, Yubin Peng, and Changqi Qu

Abstract—Direct power control (DPC) and model predictive to the grid voltage vector. Although good steady state per-
control (MPC) are two well known methods to achieve direct formance and quick dynamic response is obtained in VOC,
control of active power and reactive power in PWM rectifiers. it suffers from the tuning work of internal current controllers
They both select a voltage vector, according to a switching table
or by minimizing a cost function, and then apply it in the and relies heavily on the system parameters [6]–[8].
next control period. The use of only one voltage vector during DPC is another kind of high performance power control
one control period leads to high power ripples and variable strategy for PWM rectifiers and its has attracted wide attention
switching frequency, especially in DPC. Recently duty cycle since the birth of 1998 [9]. DPC achieves very quick response
control was proposed in DPC to achieve steady state performance with simple structure by selecting a voltage vector from a
improvement. This paper proposes an improved MPC with duty
cycle control and compared it to prior DPC with duty cycle predefined switching table. As the heuristic table is not very
control in terms of power ripple reduction, dynamic response accurate, DPC presents relatively large power ripples [8], [10],
and robustness against external load disturbance. The duration of [11].
the selected vector is determined by minimizing the active power To overcome the drawback of inaccurate vector selection in
ripple during one control period. Simulation and experimental DPC, recently MPC was proposed to obtain better steady state
results are presented to confirm the theoretical study.
performance [3], [12]–[18]. MPC is similar to DPC in that it
Index Terms—Model predictive control; direct power control; also selects and applies one voltage vector during one control
duty cycle; PWM rectifier period. However, the selected vector is not obtained from a
predefined switching table, but obtained by minimizing a cost
function. In general, the cost function is a linear combination
I. I NTRODUCTION
of active power error and reactive power error. By using the
Three-phase PWM rectifiers have been widely used in system model, the future value of active power and reactive
a variety of industrial applications owing to its advanced power for each discrete switching state can be predicted. The
features, such as sinusoidal input current, small size of input voltage vector producing minimal tracking error of power is
filter, bi-directional power flow and controllable power factor selected as the best voltage vector. It is clear that the selected
[1]–[3]. Many control methods have been proposed to achieve vector from MPC is more accurate and effective than that from
decoupled control of active power and reactive power. One the heuristic switching table in DPC due to the prediction of
of the most popular and mature control methods for PWM powers in the future. The merits of MPC are more pronounced
rectifier is voltage oriented control (VOC) [4], [5], in which in the systems with low switching frequency, high order and
the grid currents are decomposed into active power component long predictions, as shown in [19]–[21]. It is shown that, in
and reactive power component in synchronous frame attached the three-level inverter-fed direct torque control (DTC) drives,
MPC with long predictive horizons achieves more accurate
Manuscript received January 21, 2016; revised May 28, 2016; accepted control of torque and stator flux than conventional DTC
July 25, 2016. Paper 2016-IPCC-0093.R1, presented at the presented at while reducing the switching frequency significantly [22].
the 2015 IEEE Energy Conversion Congress and Exposition, Montreal,
Canada, September 20-24, 2015, and approved for publication in the IEEE Because of the merits of multiple variable control, conceptual
TRANSACTIONS ON INDUSTRY APPLICATIONS by the Industrial Power simplicity and flexibility to handle nonlinear constraints, MPC
Converter Committee of the IEEE Industry Applications Society. This work has received wide attention throughout the world [12], [20],
was sponsored in part by the National Natural Science Foundation of China
under Grant 51577003 and 51207003, in part by Beijing Natural Science [23]–[34].
Foundation under grant 3162012, and in part by Beijing Nova Program under Although DPC and MPC are different in the principle of
Grant xx2013001. vector selection, both of them apply only one voltage vector
Y. Zhang is with Inverter Technologies Engineering Research Center of
Beijing, North China University of Technology, Beijing, 100144, China, in the next control period. As a result, relatively high power
and with Collaborative Innovation Center of Electric Vehicles in Beijing, ripples and variable switching frequency can be observed.
and also with Collaborative Innovation Center of Key Power Energy-Saving Furthermore, to achieve relatively satisfactory performance,
Technologies in Beijing (email: yozhang@ieee.org).
Y. Peng and C. Qu are with Inverter Technologies Engineering Research the sampling frequencies of both DPC and MPC have to be
Center of Beijing, North China University of Technology, Beijing, 100144, high, generally tens of kilohertz. Recently the concept of duty
China (1098665228@qq.com; qcqdhr@sina.com). cycle control was introduced in DPC to achieve steady state
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. performance improvement [8], [15]. For example, the principle
Digital Object Identifier 10.1109/TIA.2016.××××××× of active power ripple minimization is employed to determine

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

ea R L
e
ea R L O b Load
ec
eb a C
ea eb ec
RL ia ib ic Udc
O b ea
ec e
c 3/2 transformation & Pulse - Udc ref
O b
ec
p, q calculation generation
e i tv
PI
Duty cycle
Delay compensation calculation
Fig. 1. Topology of a two-level three-phase AC/DC converter.
v
p k+1 pref
the duration of the selected vector from the switching table in Cost fun. Min. (MPC)/
qref
DPC [6]. qk+1 Switching table (DPC) 0
This paper makes a further step to [6] by replacing the
vector selection with the principle of cost function mini- Fig. 2. Control diagram of DPC/MPC with duty cycle control
mization in MPC. It is expected that better steady state
performance than conventional DPC/MPC and DPC with duty 8000
cycle can be achieved [35], even if relatively high sampling V4 V5
L V6 V1 V2 V3 V4
6000
ea R
frequency is used and short prediction horizon is considered.

dP/dt (kW/s)
eb
4000
A detailed comparative study of DPC and MPC with duty O Load
ec
2000
V0,7
cycle control is carried out in this paper, including steady
ea eb ec ia ib ic
state performance, dynamic response and robustness against 0 Udc
external load disturbance. The features of both methods are −2000
300 -
ref
0 3/2 transformation
60 120 & 180 Pulse
240 Udc
360
identified and confirmed by the simulation and experimental
P, Q calculation generation
results. A comprehensive conclusion is drawn, which is useful 6000
for the selection of appropriate strategy. V3 V4 V5 V6 tv V1 V
4000 PI2
sector Vector
dQ/dt (kVar/s)

II. M ODEL OF PWM R ECTIFIER 2000 duration


V0,7
0
The circuit of a three-phase AC/DC converter is shown V1 V0
in Fig. 1, where ea , eb , ec stands for grid phase voltage, L −2000
P Pref
and R are the inductance and equivalent series resistance of −4000
0 60 Q 120 Vector selection
180 240 Q ref
300
interconnecting reactors. position of grid voltage vector (in degree) 0 360
Using three-phase to two-phase transformation [5], the
mathematical model of a PWM rectifier in stationary αβ frame Fig. 3. Power slopes versus grid voltage position for various rectifier voltage
vectors (assuming p = 900 W and q = 0).
can be expressed as [13]

di III. P RINCIPLE OF DPC AND MPC W ITH D UTY C YCLE


e = Ri + L
+v (1)
dt C ONTROL
where v, e and i are rectifier voltage vector, grid voltage In conventional DPC and MPC for PWM rectifier, the
vector and grid current vector, respectively; R and L are the selected vector is applied during the whole control period,
equivalent series resistance and inductance of input line. producing irregular power ripples and rich current harmonics.
The complex power S at grid side is calculated as [36] By introducing duty cycle control, the selected vector is
applied only for a fraction of control period and the rest
3 ∗
S = p + jq =
(i e) (2) is allocated for a null vector. This brings the benefits of
2
better steady state performance and relatively fixed switching
where ∗ indicates the conjugate of a complex vector. frequency. The control diagram of DPC/MPC with duty cycle
Under the condition of sinusoidal and balanced three-phase is illustrated in Fig. 2 and the details of each block are
system, namely explained in the following texts.
de
= jωe (3)
dt
A. Vector Selection in DPC
the slope of complex power can be derived from (1) to (3) as
[6], [8] According to the equation in (4), the slopes of active power
  and reactive power versus grid voltage position for various
dS 1 3 2 
= |e| − v ∗ e − (R − jωL) · S (4) rectifier voltage vectors can be obtained and illustrated in
dt L 2 Fig. 3, where p = 900 W and q = 0 Var. The parameters
where ω is the grid frequency in rad/s. to calculate the power slope curves in Fig. 3 are listed in

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

TABLE I
ACTIVE VECTOR SELECTION FOR THE PROPOSED DPC The active power and reactive power at the next control
period can be easily predicted from (4). For two-level PWM
P Q Selected vector rectifier, it is feasible to evaluate the cost function (5) for
↑ ↑ Vk+3
↑ ↓ Vk−1 each voltage vector online, because there are only 7 different
↓ ↑ Vk+1 voltage vectors. However, the number of predictions can be
↓ ↓ Vk further reduced if the principle in [12] is employed. The best
non-zero (active) voltage vector is the one most close to the
V3(010) V2(110) position of power error vector
Sector 2
k+1 k+1
M (−S0∗ ) = −S ∗ref − (−S ∗ )0

(6)
Sector 3 Sector 1
V4(011) V1(100) where

Sector 4 Sector 6 (−S ∗ )0


k+1 tsc 
= (−S ∗ ) +
k
−1.5E 2 + (R + jωL) · (S ∗ )
k

L
(7)
Sector 5 More details regarding the principle of fast vector selection
V5(001) V6(101) can be found in [12].

Fig. 4. Rectifier voltage vectors and sector division of DPC.

C. Duty Cycle Control in DPC and MPC


Table II, which are the same as those used in simulation and
experimental results. Summarizing the results in Fig. 3, the After selecting the active vector from DPC and MPC, the
switching table for vector selection in DPC can be constructed next step is to determine the optimal duration of the selected
in Table I [10], [37], where k is the sector number and is active vector. Generally, the duration of the selected vector
cycling counted. The sector division of DPC is shown in is related to the operating points. This is easy to understand
Fig. 4. It should be noted that DPC does not select accurate in SVM-based methods such as VOC, because the reference
voltage vector in all working points. As shown in Fig. 3, voltage vector (including both modulation index and phase
when active power is decreased, the selected vector cannot angle) is directly obtained. Different from VOC, it is difficult
accurately satisfy this requirement in the vicinity of sector to analyze the average modulation index in DPC and MPC,
changing. Various switching tables have been proposed in the because only one voltage vector is selected during one control
literature [8]–[10], [38], but they cannot solve the problem period. However, both active power and reactive power can
thoroughly, because the switching table in DPC has inherent track the reference value well in DPC/MPC. This means that
drawbacks and it cannot cover every working points. In fact, the modulation index relating to operating point in DPC/MPC
the power slope is not only dependent on the applied voltage is achieved in an average sense, which is assured by the closed
vector, but is also related to specific operating points, as shown loop control and operating principle of them. In this paper, the
in (4). So, the vector selected from DPC is not always effective duty cycle control was introduced in both DPC and MPC to
but sometimes even incorrect. More details regarding vector further improve the steady state performance. The duration of
selection in DPC can refer to [10]. the selected vector is such determined that the RMS ripple
of active power during one control period is minimized. The
feasibility of introducing duty cycle control lies in that it is not
B. Vector Selection in MPC necessary to apply the selected voltage vector for the whole
Different from DPC, where the vector is selected from a control period to achieve power regulation. In fact, applying
predefined switching table, in MPC, the vector is selected by the selected vector for only a fraction of control period is
evaluating a cost function consisting of power errors for each sufficient to achieve the same power regulation, but produces
converter voltage vector. The one minimizing the cost function much less power ripples. It is possible to use other principles
is selected as the best voltage vector. It is evident the process to determine the duration of the selected vector and they
of vector selection in MPC is more accurate than DPC by should also work, as shown in [39] for a DTC drive with
taking the accurate model of system into account. The cost various duty cycle methods. The vector selection and vector
function can be expressed as the squared error of complex duration execute online in a closed-loop fashion, so the power
power [12]: regulation ability as well as the average modulation index is
not affected.
2 2
F = S ref − S k+1 = |pref − pk+1 |2 + q ref − q k+1
Suppose the active power slopes for the selected active
(5) vector and zero vector are f1 and f2 , respectively, which
can be easily obtained from the real component of (4). The
where S ref = pref + jq ref and S k+1 = pk+1 + jq k+1 are square ripple of active power during one control period can
the reference value and predicted value of complex power. be analytically expressed as

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

TABLE II
S YSTEM AND C ONTROL PARAMETERS
Z ts
1 2
F (ts ) = pk + f1 t − pref dt Line resistance R 0.3 Ω
tsc 0 Line inductance L 10 mH
Z tsc DC-bus capacitor C 840 µF
1 2
+ pk + f1 ts + f2 (t − ts ) − pref dt (8) Load resistance RL 97 Ω
tsc ts Line-line voltage (RMS) UN 150 V
Line voltage frequency f 50 Hz
where tsc and ts are control period and the duration of the DC voltage Udc 300 V
active vector.
Minimizing the active power ripple is equivalent to mini-
mizing F (ts ). This can be achieved by solving the equation The equations for predicting pk+2 and q k+2 are similar to
dF (ts )/dt = 0. The resulting optimal duration of the active the equations in (10) and (11), which are expressed as
vector is calculated as


2 pref − pk − f2 tsc k+2 k+1 3 h k+1 2 i
− Re conj v k+1 · ek+1

p =p + e
ts = (9) 2L
2f1 − f2 
R k+1 k+1
It should be noted that the value of ts is set to zero if it is − p − ωq tsc (13)
L
negative and to tsc if it is bigger than tsc . 
−3
q k+2 = q k+1 + Im conj v k+1 · ek+1
 
2L
D. Vector Sequence 
R k+1 k+1
After obtaining the selected voltage vector following the − q + ωp tsc (14)
L
principle introduced in Section III-B, it will be applied for an
optimal duration according to (9) during one control period. where the grid voltage vector at (k + 1)th instant is predicted
It should be noted that the active vector should be followed from (3) as
by an appropriate zero vector with minimal switching jumps. ek+1 = ejωtsc ek (15)
For vectors (110, 011, 101), they should be followed by
(111), while for other vectors (100, 010, 001), they should IV. S IMULATION AND E XPERIMENTAL R ESULTS
be followed by (000). Hence, no more than one commutation To confirm the effectiveness of the proposed MPC with duty
event occurs simultaneously during one control period. cycle control, both digital simulation and experimental tests
are carried out on a two-level three-phase PWM rectifier. The
E. Delay Compensation results obtained from prior DPC with duty cycle control [6]
It is well known in digital control system that, the actual are also presented for the aim of comparison. The sampling
voltage vector will not be applied until the next control period frequencies for the proposed DPC and MPC with duty cycle
due to the updating mechanism of PWM block. Hence, to control are both 20 kHz and the system and control parameters
mitigate the influence of one-step delay, the state variables are listed in Table II. Although the sampling frequency is
at (k + 1)th instant should be employed to accomplish the high, the average switching frequency is moderate, as shown
control algorithm. In other words, in DPC we should compare in the experimental results. It is possible to operate at lower
pk+1 and q k+1 with their respective reference to select the switching frequency such as 5 to 10 kHz. However, the steady
appropriate vector from the switching table. The predicted state performance would be also degraded, because the average
value of active power and reactive power are expressed as: switching frequency is very low. Hence, for the proposed two
methods, it is recommended to use high sampling frequency
 to achieve good steady state performance. The overall control
k+1 k 3 h k 2 i
e − Re conj v k · ek

p =p + diagram of DPC and MPC is illustrated in Fig. 2, where
2L the digital delay compensation has been considered for both

R methods, as introduced in Section III-E.
− pk − ωq k tsc (10)
L

−3 A. Simulation Results
q k+1 = qk + Im conj v k · ek
 
2L
 Fig. 5 presents the simulated responses to external load
R k k disturbance for the proposed DPC and MPC with duty cycle
− q + ωp tsc (11)
L control at 20 kHz sampling frequency. From top to bottom, the
Different from in DPC, in MPC the prediction of pk+1 and curves shown in Fig. 5 are active power and reactive power
q k+1
is only the first step. To compensate the one-step digital (with reference), dc-bus voltage and one-phase grid voltage
delay in MPC, the predicted active power and reactive power and grid current. An external load is applied to the system
in the cost function (5) should be replaced by pk+2 and q k+2 . at t=0.05s. During the dynamic process, the reactive power is
As a result, the cost function should be changed to maintained at zero to achieve unity power factor. It is seen
2 2 that the active power increases quickly to balance the increase
F = pref − pk+2 + 0 − q k+2 (12) in the active power. There is a small voltage drop of around 5

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

1500 1500
P&Q (VA)

1000 P
1000 Q

P&Q (VA)
500
0 500
-500
0 0.05 0.1 0.15 0
305
-500
Udc/ V

300 0 0.05 0.1 0.15


P
295 Q
10
290
100
5
10

Ua (V))

ia (A))
100 0 0
5
Ua (V))

ia (A))
0 0
-5
-5 -100
-100
-10 -10
0 0.05 0.1 0.15 0 0.05 0.1 0.15
t/s t/s
(a) (a)

1500 1500
P&Q (VA)

1000 P
1000 Q

P&Q (VA)
500
0 500
-500
0 0.05 0.1 0.15 0
305
-500
Udc/ V

300 0 0.05 0.1 0.15


P
295 Q
10
290
100
5
10
Ua (V))

ia (A))
100 0 0
5
Ua (V))

ia (A))

0 0
-5
-5 -100
-100
-10 -10
0 0.05 0.1 0.15 0 0.05 0.1 0.15
t/s t/s
(b) (b)
Fig. 5. Simulated responses to external load disturbance for (a) the proposed Fig. 6. Simulated responses under the condition of stepped changes in active
DPC with duty cycle control at 20 kHz sampling frequency and (b) the power reference for (a) the proposed DPC with duty cycle control at 20 kHz
proposed MPDPC with duty cycle control at 20 kHz sampling frequency. sampling frequency and (b) the proposed MPDPC with duty cycle control at
20 kHz sampling frequency.
vector no. (DPC)

6
V in the dc-bus voltage and it recovers to its reference value
quickly, which confirms the robustness of both methods in 4
rejecting external load disturbance. The dynamic responses of
both methods are very similar. However, it is clear that the 2

proposed MPC with duty cycle control presents much lower 0


power ripple and less current harmonics, especially in the 0 0.05 0.1 0.15
reactive power ripple. t/s

The results in Fig. 5 was obtained under the condition of


vector no. (MPC)

constant dc bus voltage, where the active power was generated 6

from an outer dc voltage loop. Fig. 6 further shows the 4


simulated responses under the condition of stepped changes
in active power reference for both methods, where the active 2
power reference is directly set rather than from the outer dc
0
voltage controller. The active power reference steps from 600 0 0.05 0.1 0.15
W to 1000 W at t = 0.05 s and then steps down to 600 W t/s
at t = 0.1 s. From top to bottom, the curves shown in Fig. 6
are active power and reactive power, one phase grid voltage Fig. 7. Simulated selected non-zero vector for DPC and MPC under the
condition of stepped changes in active power reference.
and grid current. It is clearly seen that in both methods, the
active power and reactive power tracks the reference value
very quickly, showing similar dynamic performance. However,

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

5
ia /A

-5

0 0.01 0.02 0.03 0.04


t/s
Fundamental ( 50Hz ) = 5.564, THD = 5.4577%
4
4
3
Hn /H1 (%)

2
2
0
0 10 20 30
1

0
0 50 100 150 200 250 300 350 400 450 500
Harmonic order
Fig. 9. Experimental setup of two-level PWM rectifier.
(a)

5
control. On the contrary, the low order harmonics in MPC with
duty cycle control are much reduced. The existence of low
ia /A

0
order harmonics in DPC mainly comes from the predefined
switching table, which is not always effective but sometimes
-5
even incorrect. There are some other reasons contributing to
0 0.01 0.02 0.03 0.04 the low order harmonics in both methods, such as dead time,
t/s
inaccurate system parameters, AD accuracy and unmodeled
Fundamental ( 50Hz ) = 5.3272, THD = 2.4801%
4
4
factors, etc. In both methods, the current harmonics concen-
3
trate on the sampling frequency of 20 kHz, which provides
Hn /H1 (%)

2 some convenience for filter design. This is mainly caused by


2 the application of both an non-zero vector and a zero vector
0
0 10 20 30 during one control period.
1

0
0 50 100 150 200 250 300 350 400 450 500
Harmonic order B. Experimental Tests
(b)
Fig. 8. Simulated harmonic spectrum of grid current at p = 1000 W and Apart from the simulation results, experimental tests are also
q = 0 Var for (a) the proposed DPC with duty cycle control at 20 kHz
sampling frequency and (b) the proposed MPDPC with duty cycle control at
carried out on a two-level PWM rectifier. The experimental set-
20 kHz sampling frequency. up is illustrated in Fig. 9, where a floating point digital signal
processor (DSP) TMS320F28335 is employed to accomplish
the two control methods. During the experiments, all variables
the prior DPC with duty cycle control presents higher and are displayed and recorded using a digital oscilloscope via on-
irregular power ripples, especially in the reactive power. On board DA converter except the grid current, which is measured
the contrary, the power ripples in the proposed MPC with duty directly by a current probe.
cycle control are much smaller and the grid current are also Firstly, the steady state responses of both methods are pre-
more sinusoidal in shape. The selected active vectors for both sented in Fig. 10, where the active power reference is 1000 W
methods are shown in Fig. 7, which are obtained under the and the reactive power reference is zero to achieve unit power
same condition as Fig. 6. It is clearly seen that the vector factor. From top to bottom, the curves shown in Fig.10 are
selected from DPC is irregular and arbitrary. On the contrary, active power reference, active power, reactive power and one
the vector selected from MPC exhibits a regular behavior. phase grid current. It can be seen that the proposed MPC with
The results clearly show that the vector selected from MPC duty cycle control presents better steady state performance
is significantly different from that from DPC, and is more than proposed DPC with duty cycle control, especially in the
effective in reducing power errors. reactive power ripples. A quantitative harmonic analysis of
Fig. 8 illustrates the harmonic spectrum of both methods. active power and reactive power for both methods is illustrated
It is seen that the current THD of the proposed MPC with in Fig. 11. It is seen that most of the low order harmonics
duty cycle control is only 2.48%, while the current THD of of active power in MPC is lower than those of DPC except
prior DPC with duty cycle is as high as 5.46%. It should the 300 Hz harmonics. For the reactive power, the low order
be noted that there are some low order (even and odd) harmonics of DPC are much bigger than those of MPC,
harmonics in both methods, especially in DPC with duty cycle especially the 300 Hz and 600 Hz harmonics.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

1100

P ref [400W/div] 1050

P (W)
P [400W/div] 1000

950
Q [400Var/div]
900
0 0.01 0.02 0.03 0.04
ia [10A/div] time (s)

Mag (% of mean value of P)


1 DPC MPC

(a)
0.5

P ref [400W/div]
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
P [400W/div]
frequency (Hz)
(a)
Q [400Var/div]
200

ia [10A/div]
100

Q (Var) 0

-100
(b)
Fig. 10. Steady state response of (a) DPC with duty cycle control, (b) -200
proposed MPDPC with duty cycle control. 0 0.01 0.02 0.03 0.04
time (s)
Mag (% of mean value of P)

6 DPC MPC

2
At steady state of p=1000 W and q=0 Var, the average
switching frequencies of the presented DPC with duty cycle 0
and MPC with duty cycle control are 8.71 kHz and 8.82 0 200 400 600 800 1000 1200 1400 1600 1800 2000
frequency (Hz)
kHz, respectively, which are similar to each other and well
below the sampling frequency of 20 kHz. Reducing the (b)
sampling frequency can achieve even lower average switching Fig. 11. Harmonic spectrum of active power and reactive power for (a) DPC
with duty cycle control, (b) proposed MPDPC with duty cycle control.
frequency, but the steady state performance would also be
affected seriously, which is unsuitable for DPC/MPC branched
methods. standard deviation function [39], which is expressed as
v
u N N
!2
The grid current quality of the proposed MPC is also better u1 X 1 X
than that of DPC with duty cycle control, which is confirmed prip = t pi − pi (16)
N i=1 N i=1
by the harmonic spectrum shown in Fig. 12. It is seen that the v
current THD of the proposed MPC is only 2.88%, which is u N N
!2
u1 X 1 X
much lower than the value of 5.77% in the prior DPC with qrip = t qi − qi (17)
duty cycle control. For grid connected applications, the current N i=1 N i=1
THD is generally limited to below 5% [40]. So, the proposed where N is the sampling number of active power and reactive
MPC with duty cycle control well respect the grid codes, while power during a short period of 0.1s. It is seen from Fig. 13 that
the DPC with duty cycle control fails, even if the duty cycle the proposed MPC exhibits much lower power ripples than
control is introduced. In both methods, the current harmonic DPC at both 600 W and 1000 W active power. The power
concentrates on the multiples of 10 kHz, which is similar to ripple reduction is very evident especially in the light load of
SVM-based methods and provides some convenience for the 600 W active power.
filter design. Secondly, the dynamic responses of both methods are com-
pared. Fig. 14 presents the experimental waveform when the
A quantitative comparison of active power and reactive reference value of active power is changed from 600 to 1000
power ripples for both methods at various operating points W suddenly. The reactive power is kept at zero to achieve
is illustrated in Fig. 13. The power ripple is calculated by the unit power factor. It can be seen that both method achieve

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

active power ripple (W)


30

5 20
Ia /A

0 10

-5 0
0 0.01 0.02 0.03 0.04 VA
t/s DPC MPC

reactive power ripple (Var)


Fundamental ( 50Hz ) = 5.3435, THD = 5.7688% 60
4 4

3 40
H n /H1 (%)

2
2

0 20
1
0 10 20 30
0 0
0 50 100 150 200 250 300 350 400 450 500 600 1000
Harmonic order VA
(a)
Fig. 13. Comparisons of active power ripple and reactive power ripple
for DPC with duty cycle control and the proposed MPDPC with duty cycle
5 control.
Ia /A

-5
P ref [400W/div]
0 0.01 0.02 0.03 0.04 P [400W/div]
t/s
Fundamental ( 50Hz ) = 5.4824, THD = 2.8828%
4 4
Q [400Var/div]
3
H n /H1 (%)

2
2 ia [10A/div]

1 0
0 10 20 30
0 (a)
0 50 100 150 200 250 300 350 400 450 500
Harmonic order
(b)
Fig. 12. Harmonic spectrum of grid current at p = 1000 W and q = 0 Var P ref [400W/div]
for (a) DPC with duty cycle control, (b) proposed MPDPC with duty cycle P [400W/div]
control.

Q [400Var/div]
decoupled control of active power and reactive power and
they have very similar quick dynamic response. However, the
ia [10A/div]
reactive power ripple in the proposed MPC is much smaller
than that in DPC.
Finally, the robustness against load disturbance is shown
in Fig. 15, where the dc-bus voltage is maintained at 300 V. (b)
An external load of 1000 W is suddenly applied to the PWM Fig. 14. Transient response to step change in active power, (a) DPC with
rectifier and the active power increases quickly to balance the duty cycle control, (b) proposed MPDPC with duty cycle control.
load change. From top to bottom, the curves shown in Fig. 5
are active power, reactive power, dc voltage and grid current.
There is very insignificant voltage drop, confirming the ro- predicting the power error vector caused by zero vector only
bustness of DPC and MPC against external load disturbance. for once, which is much more efficient than the conventional
The dynamic responses of both methods are very similar, but enumeration-based MPC. The duration of the selected active
the proposed MPDPC has smaller power ripples and the grid vector is obtained based on the principle of minimizing the
current is more sinusoidal with less harmonics. active power ripple during one control period. The proposed
MPC with duty cycle control is compared to prior DPC
V. C ONCLUSION with duty cycle control by evaluating their performance in
This paper proposes an improved MPC with active power terms of steady state performance, dynamic response and
ripple minimization. The active vector is selected quickly by robustness against external load disturbance. The simulation

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

[9] T. Noguchi, H. Tomiki, S. Kondo, and I. Takahashi, “Direct power


control of pwm converter without power-source voltage sensors,” IEEE
Trans. Ind. Appl., vol. 34, no. 3, pp. 473–479, 1998.
P [1000W/div] [10] Y. Zhang, C. Qu, Z. Li, and Y. Zhang, “Mechanism analysis and
Q [1000W/div]
experimental study of table-based direct power control,” in Proc. Int.
Conf. Electrical Machines and Systems ICEMS 2013, 2013, pp. 2213–
U dc [300V/div] 2218.
[11] Y. Zhang and C. Qu, “Direct power control of a pulse width modulation
rectifier using space vector modulation under unbalanced grid voltages,”
ia [10A/div]
IEEE Trans. Power Electron., vol. 30, no. 10, pp. 5892–5901, 2015.
[12] Y. Zhang and W. Xie, “Low complexity model predictive control—single
vector-based approach,” IEEE Trans. Power Electron., vol. 29, no. 10,
pp. 5532–5541, 2014.
[13] P. Cortes, J. Rodriguez, P. Antoniewicz, and M. Kazmierkowski, “Direct
(a) power control of an afe using predictive control,” IEEE Trans. Power
Electron., vol. 23, no. 5, pp. 2516 –2523, sept. 2008.
[14] Y. Zhang, W. Xie, Z. Li, and Y. Zhang, “Low-complexity model
predictive power control: Double-vector-based approach,” IEEE Trans.
P [1000W/div] Ind. Electron., vol. 61, no. 11, pp. 5871–5880, 2014.
[15] Y. Zhang, Y. Peng, and H. Yang, “Performance improvement of two-
Q [1000W/div] vectors-based model predictive control of PWM rectifier,” IEEE Trans.
U dc [300V/div] Power Electron., vol. 31, no. 8, pp. 6016–6030, 2016.
[16] D. E. Quevedo, R. P. Aguilera, M. A. Perez, P. Cortes, and R. Lizana,
“Model predictive control of an AFE rectifier with dynamic references,”
ia [10A/div] IEEE Trans. Power Electron., vol. 27, no. 7, pp. 3128–3136, 2012.
[17] Y. Zhang, W. Xie, Z. Li, and Y. Zhang, “Model predictive direct power
control of a PWM rectifier with duty cycle optimization,” IEEE Trans.
Power Electron., vol. 28, no. 11, pp. 5343–5351, 2013.
[18] Y. Zhang, C. Qu, and J. Gao, “Performance improvement of direct power
(b) control of pwm rectifier under unbalanced network,” IEEE Trans. Power
Electron., vol. PP, no. 99, pp. 1–1, 2016.
Fig. 15. Responses to external disturbance in load for (a) DPC with duty [19] J. Holtz, “Advanced pwm and predictive control—an overview,” IEEE
cycle control, (b) proposed MPDPC with duty cycle control. Trans. Ind. Electron., vol. 63, no. 6, pp. 3837–3844, June 2016.
[20] T. Geyer and D. Quevedo, “Performance of multistep finite control set
model predictive control for power electronics,” IEEE Trans. Power
Electron., vol. 30, no. 3, pp. 1633–1644, 2015.
and experimental results prove that, with the same sampling
[21] J. Scoltock, T. Geyer, and U. Madawala, “Model predictive direct power
frequency, the proposed MPC can achieve reduced power control for grid-connected npc converters,” IEEE T. Ind. Electron.,
ripples and lower current THD while maintaining similar vol. 62, no. 9, pp. 5319–5328, Sept 2015.
dynamic response to DPC with duty cycle control. Hence, [22] T. Geyer, G. Papafotiou, and M. Morari, “Model predictive direct torque
control -part I: Concept, algorithm, and analysis,” IEEE Trans. Ind.
MPC with duty cycle control is a powerful strategy even Electron., vol. 56, no. 6, pp. 1894 –1905, june 2009.
in the low power applications with relatively high switching [23] R. Aguilera, P. Lezana, and D. Quevedo, “Finite-control-set model
frequency, low system order and short prediction horizon. predictive control with improved steady-state performance,” IEEE Trans.
Ind. Informat., vol. 9, no. 2, pp. 658 –667, may 2013.
[24] A. Damiano, G. Gatto, I. Marongiu, A. Perfetto, and A. Serpi, “Op-
R EFERENCES erating constraints management of a surface-mounted pm synchronous
machine by means of an fpga-based model predictive control algorithm,”
[1] Y. Zhang and C. Qu, “Table-based direct power control for three-phase IEEE Trans. Ind. Informat., vol. 10, no. 1, pp. 243–255, Feb 2014.
AC/DC converters under unbalanced grid voltages,” IEEE Trans. Power [25] Y. Zhang and H. Yang, “Model predictive torque control of induction
Electron., vol. 30, no. 12, pp. 7090–7099, 2015. motor drives with optimal duty cycle control,” IEEE Trans. Power
[2] J. Rodriguez, J. Dixon, J. Espinoza, J. Pontt, and P. Lezana, “Pwm Electron., vol. 29, no. 12, pp. 6593–6603, 2014.
regenerative rectifiers: state of the art,” IEEE Trans. Ind. Electron.,
[26] H. Young, M. Perez, J. Rodriguez, and H. Abu-Rub, “Assessing finite-
vol. 52, no. 1, pp. 5 – 22, feb. 2005.
control-set model predictive control: A comparison with a linear current
[3] Y. Zhang and C. Qu, “Model predictive direct power control of PWM
controller in two-level voltage source inverters,” IEEE Ind. Electron.
rectifiers under unbalanced network conditions,” IEEE Trans. Ind. Elec-
Mag., vol. 8, no. 1, pp. 44–52, 2014.
tron., vol. 62, no. 7, pp. 4011–4022, 2015.
[4] M. Malinowski, M. P. Kazmierkowski, and A. M. Trzynadlowski, [27] R. Ramirez, J. Espinoza, F. Villarroel, E. Maurelia, and M. Reyes, “A
“A comparative study of control techniques for pwm rectifiers in ac novel hybrid finite control set model predictive control scheme with
adjustable speed drives,” IEEE Trans. Power Electron., vol. 18, no. 6, reduced switching,” IEEE Trans. Ind. Electron., vol. 61, no. 11, pp.
pp. 1390–1396, 2003. 5912–5920, 2014.
[5] V. Blasko and V. Kaura, “A new mathematical model and control [28] Y. Zhang and H. Yang, “Generalized two-vector-based model-predictive
of a three-phase ac-dc voltage source converter,” IEEE Trans. Power torque control of induction motor drives,” IEEE Trans. Power Electron.,
Electron., vol. 12, no. 1, pp. 116–123, 1997. vol. 30, no. 7, pp. 3818–3829, 2015.
[6] Y. Zhang, C. Qu, Z. Li, and W. Xu, “An improved direct power control [29] S. Kwak and J.-C. Park, “Switching strategy based on model predictive
of PWM rectifier with active power ripple minimization,” in Energy control of vsi to obtain high efficiency and balanced loss distribution,”
Conversion Congress and Exposition (ECCE), 2014 IEEE, 2014, pp. IEEE Trans. Power Electron., vol. 29, no. 9, pp. 4551–4567, 2014.
527–533. [30] Z. Ma, S. Saeidi, and R. Kennel, “Fpga implementation of model
[7] L. Hang, S. Liu, G. Yan, B. Qu, and Z. LU, “An improved deadbeat predictive control with constant switching frequency for pmsm drives,”
scheme with fuzzy controller for the grid-side three-phase pwm boost IEEE Trans. Ind. Informat., vol. 10, no. 4, pp. 2055–2063, Nov 2014.
rectifier,” IEEE Trans. Power Electron., vol. 26, no. 4, pp. 1184–1191, [31] Y. Zhang and H. Yang, “Model-predictive flux control of induction
2011. motor drives with switching instant optimization,” IEEE Trans. Energy
[8] Y. Zhang, Z. Li, Y. Zhang, W. Xie, Z. Piao, and C. Hu, “Performance Convers., vol. 30, no. 3, pp. 1113–1122, 2015.
improvement of direct power control of pwm rectifier with simple [32] Y. Zhang, H. Yang, and B. Xia, “Model predictive control of induction
calculation,” IEEE Trans. Power Electron., vol. 28, no. 7, pp. 3428– motor drives: Torque control versus flux control,” IEEE Trans. Ind.
3437, July 2013. Appl., vol. 52, no. 5, 2016.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2596240, IEEE
Transactions on Industry Applications

10

[33] L. Tarisciotti, P. Zanchetta, A. Watson, J. Clare, M. Degano, and Yongchang Zhang (M’10) received the B.S. degree
S. Bifaretti, “Modulated model predictive control for a three-phase active from Chongqing University, China, in 2004 and the
rectifier,” IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1610–1620, 2015. Ph.D. degree from Tsinghua University, China, in
[34] Y. Zhang and H. Yang, “Two-vector-based model predictive torque 2009, both in electrical engineering.
control without weighting factors for induction motor drives,” IEEE From August 2009 to August 2011, he was a
Trans. Power Electron., vol. 31, no. 2, pp. 1381–1390, 2016. Postdoctoral Fellow at the University of Technology
[35] Y. Zhang, Y. Peng, and C. Qu, “Comparative study of model predictive Sydney, Australia. He joined North China Univer-
control and direct power control for PWM rectifiers with active power sity of Technology in August 2011 as an associate
ripple minimization,” in Energy Conversion Congress and Exposition professor. Currently he is a full professor and the
(ECCE), 2015 IEEE, 2015, pp. 3823–3830. vice director of Inverter Technologies Engineering
[36] H. Akagi, Y. Kanazawa, and A. Nabae, “Instantaneous reactive power Research Center of Beijing. He has published more
compensators comprising switching devices without energy storage than 100 technical papers in the area of motor drives, pulsewidth modulation
components,” IEEE Trans. Ind. Appl., vol. 20, no. 3, pp. 625–630, 1984. and AC/DC converters. His current research interest is model predictive
[37] A. Bouafia, J.-P. Gaubert, and F. Krim, “Analysis and design of new control for power converters and motor drives.
switching table for direct power control of three-phase pwm rectifier,”
in Proc. 13th Power Electronics and Motion Control Conf. EPE-PEMC
2008, 2008, pp. 703–709.
[38] A. Baktash, A. Vahedi, and M. A. S. Masoum, “Improved switching
table for direct power control of three-phase pwm rectifier,” in Proc. Yubin Peng was born in 1990. He received the
Australasian Universities Power Engineering Conf. AUPEC 2007, 2007, B.S. degree from Beijing Information Science and
pp. 1–5. Technology University in 2013, and the master’s
[39] Y. Zhang and J. Zhu, “Direct torque control of permanent magnet syn- degree from the North China University of Tech-
chronous motor with reduced torque ripple and commutation frequency,” nology, Beijing, China, in 2015, both in electrical
IEEE Trans. Power Electron., vol. 26, no. 1, pp. 235 –248, jan. 2011. engineering.
[40] IEEE Recommended Practices and Requirements for Harmonic Control His research interest is model predictive control
in Electrical Power Systems, IEEE Std. 519-1992, 1993. of three-level PWM rectifiers.

Changqi Qu was born in 1988. He received the


B.S. degree from Beijing Jiaotong University, Bei-
jing, China, in 2009, and the master’s degree from
the North China University of Technology, Beijing,
China, in 2015, both in electrical engineering.
His research interest is control of PWM rectifiers
under unbalanced grid voltages.

0093-9994 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

S-ar putea să vă placă și