You are on page 1of 8

Optimal Space Vector Modulation Control for

Three-Phase Inverter

Su Xiaofang1, Rao Binbin2,*, Li Shijie2, and Zheng Ruilan2


1
Hubei University of Education, China
2
School of Electrical Engineering,
Wuhan University,Wu Han, China
yiyantangbin@163.com

Abstract. This paper proposes a method for optimization of the harmonic per-
formance of three-Phase inverter under Space Vector Modulation(SVM) control.
It is shown that SVM provides a number of degrees of freedom, which make it
suitable for optimization, subject to desired constraints. An immune algorithm
(IA) optimizer is used to solve the optimization problem. Simulation results are
shown to result in significant reduction of dominant harmonics compared with
the traditional control strategies.

Keywords: Immune Algorithm; space vector modulation; degrees of freedom;


Three-Phase Inverter; harmonic analysis.

1 Introduction
With the rapid development of power industry, The application of inverter is more and
more widely. High power electronic devices are being used increasingly to control and
facilitate flow of electric power while meeting stringent operating conditions of today’s
heavily loaded networks. One of such devices is voltage-sourced converter (VSC) that
acts as a controlled voltage source, converting a dc voltage to an ac voltage with desired
frequency, phase, and magnitude [1-7]. Fig. 1 shows the schematic diagram of a
two-level VSC.
The output waveform of the VSC contains harmonics, largest of which occurs
around the switching frequency.
Sinusoidal PWM (SPWM) has a relatively robust harmonic spectrum, i.e., the har-
monic spectrum of the resulting waveform is tied to the selected switching frequency. In
contrast, space vector modulation (SVM), which is a relatively new approach to wave-
form synthesis using a VSC, offers several degrees of freedom that can be used effec-
tively to design an improved harmonic spectrum and obtain the desired waveform
quality.
SVM has found numerous applications in both power system schemes, such as static
compensator (STATCOM) [8] and high voltage direct current converter systems [9], and
electrical drive applications [10]. In [3], An optimized space vector modulation se-
quence for improved harmonic performance is presented.

*
Corresponding author.

M. Ma (Ed.): Communication Systems and Information Technology, LNEE 100, pp. 609–616.
springerlink.com © Springer-Verlag Berlin Heidelberg 2011
610 S. Xiaofang et al.

Fig. 1. Schematic diagram of a two-level VSC

This paper aims to optimize space vector sequences through exploiting the degrees
of freedom available in SVM. The objective of the optimization is to relieve the filtering
requirement by distributing the energy of the waveform more uniformly among har-
monic components. An overview of the SVM methods is provided in the next section,
which also describes the parameters that are used in the optimization.
As will be described later, the problem involves both real and integer variables.
Therefore, Immune Algorithms (IA), which can handle such mixed-integer problems,
are used for its optimization. IA have been previously applied to find the filter pa-
rameters in electrical drives [3-7], as well as to a two level SPWM converter to minimize
the harmonic content of its output waveform.

2 SVM Methods and the Degrees of Freedom


In conventional SPWM, which is an analog domain method, the duration of each pulse
is found through comparison of a sinusoidal reference waveform and a triangular carrier
waveform. A digital domain variation of PWM, which is the SVM, on the other hand,
directly computes the duration of voltage pulses using the amplitude and angular loca-
tion of the reference vector.
In a three-phase two level VSC, the output voltage of a leg can be either +Vdc/2
or−Vdc/2 (see Fig. 1), and as such, the VSC can be placed in eight states, depending on
the ON/OFF status of its six switches (switches on the same leg have complementary
states). The voltages corresponding to each state can be transformed to a space vector
using six of these space vectors, denoted by active vectors V1 to V6 in Fig. 2, point to
the vertices of a hexagon; two of the states translate to space vectors with zero length and
are denoted by zero vectors V0 and V7 located at the origin[3-4].
Similarly, any set of balanced reference phase voltages va,vb and vc can be repre-
sented as a corresponding reference space vector V in two dimensions using function
(1).
2
V = (va + vb e j 2π /3 + vc e − j 2π /3 ) (1)
3
Optimal Space Vector Modulation Control for Three-Phase Inverter 611

Any reference vector that lies entirely within the hexagon can be decomposed to
space vectors V0 to V7, A reference vector can be decomposed using virtually any
subset of the eight space vectors. Nonetheless, decomposition is done typically using
only the two adjacent active vectors. as shown in Fig. 2, for a Vref in the first sector of
the hexagon. This is achieved by constructing voltages that average to the reference
vector , which is sampled at a rate determined by a given sampling frequency Fs.

Fig. 2. Space vectors and decomposition of Fig. 3. Conventional SVM sequence


a reference vector

The voltages are synthesized by placing the converter in respective states for desig-
nated time shares within each sampling period. The time shares are proportional to the
length of the projected vectors. The rest of the sampling period is filled with zero space
vectors V0 and/or V7. Fig.3 shows the conventional SVM sequence.
0
For example, if ∆θ=2 , one period is divided into 180, i.e., it has 180 periods of
vectors, so the sample period is

Ts = 0.02 s / 180 ≈ 111us (2)

In sector I, time shares are calculated as

T4 = 3 2 mTs sin(π 3 − θ ) (3)

T6 = 3 2 mTs sin θ (4)


Tzero = T0 + T7 = Ts − T4 − T6 (5)

where T4, T6, T0, and T7 are the time shares of the respective voltage vectors (T1 for the
first active vector V4, T6 for V6, T0 for V0, and T7 for V7), Ts is the sampling period
and is equal to 1/Fs, θ is the angle between the reference vector and the space vector V4,
and m is the modulation index, which is defined as

m = Vref (Vdc 2) (6)


612 S. Xiaofang et al.

Note that the SVM strategy offers several degrees of freedom, they are as follows .
1) It determines the time shares of space vector, but the individual time shares of T0
and T7 are not specified.
2) It does not identify the order in which they are applied. For example, the order of
vector can be V0V4V6V7, but it also can be V4V0V6V7. this can be exploited as a
degree of freedom for crafting waveforms with given properties.
3) In different period, the individual time shares of T0 and T7 can be different. For
example, if each time of T0 and T7 is applied for half of the total zero-vector time , then
it can be applied as that T0-share is 80% of the total zero-vector time for the next pe-
riod. In different period, the order in which they are applied also can be different. For
example, it can be V0V4V6V7, but it can be V4V0V6V7 in the next period.

3 SVM Optimization Based on IA

Immune Algorithm (IA) belong to a group of evolutionary algorithms that are inspired
by natural processes. Such methods seek to maximize the fitness of a population or, as in
the common notion, minimize the associated cost (or objective) function[5-7].
In IA, the solution space is represented by a generation of candidate solutions, called
chromosomes. An objective function (OF) is calculated to reflect the fitness of each
chromosome. By exploiting the degrees of freedom built-in to SVM, appropriate
switching sequences can be devised to obtain the desired waveform quality, through the
definition of an appropriate OF.

A. Objective Function
This paper is desired to lower the harmonic content of the output current, so it use
weighted THD (WTHD) as the OF. The WTHD, defined as follows, is calculated in the
same way as THD, but each harmonic component is divided by its order, so that higher
order harmonics receive lower weight and contribute less in this figure of merit.

N
Vn 2

n = 2,3,"
(
n
)
(7)
WTHD =
V1

B. Flow Of IA
The Implementation of the IA contains the following steps, the flowchart of which is
shown in Fig.4.
a) Step 1. Individual encoding and initial populations. This paper presents an ame-
liorative genetic algorithm based on integer encoding strategy for optimization problems
because of the shortages of binary encoding . To demonstrate how sequences are en-
coded, consider the conventionally used sequence V0V4V6V7. Encoding results in a
string of 0(V0(000)), 4(V4(100)), 6(V6(110)), 7(V7(111)),so the conventional SVM
sequence(0-π/6) can be encoded as Figure 5.
Optimal Space Vector Modulation Control for Three-Phase Inverter 613

Fig. 4. Flow chart of immune algorithm Fig. 5. Individual encoding

b) Sstep2. Vaccine inoculate. It can be likely to increase affinity of the antibody.


According to the practical experience or normal control strategies, some feasible solu-
tions of the practical problems can be gain. Those solutions are called as vaccine. And
those vaccines are injected into the initial generation, which can greatly quicken con-
vergence of the algorithm.
c) Step 3. Fitness calculation. IA seek to maximize the fitness of a population, and
the fitness is associated with Objective Function (OF). The definition of the OF in this
paper is to minimize the THD of the output waveform under the method of SVM.
d) Step 4. Immune selection. IA uses selection operator to simulate natural evolution.
Some antibodies of high affinity should be inherited to the next generation with greater
probability.
e) Step 5. Crossover. Members of the antibodies are paired together. The process of
crossover is that various antibodies exchange their substrings sections to generate new
antibodies. Fig.6 shows the crossover option.

ChromA 046704670467... 046747606407...


Crossoverpoint

ChromB 406747606407... 406704670467...


Crossoverpoint

Fig. 6. Crossover option Fig. 7. Inversion option

f) Step 6. Mutation. Mutation operation is a process of randomly changing one or


some alleles of an antibody. Based on the predetermined mutation rate, some antibodies
randomly selected are mutated. In this paper mutation option is for the individual time
shares of T0.
g) step 7. Inversion. Inversion operation is that vectors exchange each other during
the same period. Fig.7 shows the inversion option.
614 S. Xiaofang et al.

h) Step 8. Random addition and New generation. Some antibodies randomly created
will be injected into every generation, which is called as random addition in this paper.
This process can avoid falling into the trap of local optima.
After IA carries out above 8 operations, it will achieve the latest generation, If the
termination criterion was satisfied, the procedures end, otherwise, it will continue from
the beginning.

4 Discussion of Results

The IA optimizer is run with the aforementioned OF. Although mutation is implemented
in the IA solver as a means to reduce the likelihood of converging to a local minimum,
the stochastic nature of IA necessitates running the optimizer several times to further
reduce this possibility. For each run, the best sequence of vectors and the share of V0 are
recorded.
The main parameters used by IA in simulations were set as Table 1. The modulation
index is 0.8 and the normalized sampling Period is 111us. The DC voltage is 25V, and
the inductance is 63.7mH, the output current is 0.72A in theory.
Table 2 shows the optimized SVM sequence (0-π/3).The T0-share in different period
has also been optimized. Both the conventional and optimized SVM sequences are
forward, however, the optimized SVM positions the vectors in a different order and
allots different inactive time interval to the zero vector Z0.

Table 1. Value of the IA parameters Table 2. The optimized Svm sequence

IA PARAMETERS VALUE 047604760476046704674607640


Maximum iteration times 2000 740670467046704674067406740
764067406740674067406740670
Population Size 100
467047640674067406740674607
Population Length 240 460746076047602760276027062
Crossover probability 0.7 702760267026702670267206702
Mutation probability 0.005 672067206720672067206720672
607607260276027602760276027
Inversion probability 0.015 602760276027620720672067
Immune selection probability 0.01

Fig.8 shows the harmonic spectrum of the voltage waveform of the conventional
SVM, and Fig.9 shows the harmonic spectrum of the voltage waveform of the optimized
SVM. Through the Fig 8. and Fig 9, for both conventional SVM and conventional SVM,
harmonic components 3th, 5 th and 7th are the largest contributors to the harmonic
distortion. However, optimization has reduced the 3th, 5th and 7th harmonics. Optimi-
zation has reduced the harmonics from 0.46% and for the conventional SVM sequence
to 0.39% for the optimized sequence. What’s more, the optimization also reduces the
DC component of the output current.Such significant reductions will result in smaller
harmonic currents in the filter and thus allow the use of smaller filters.
Optimal Space Vector Modulation Control for Three-Phase Inverter 615

Fig. 8. Harmonic spectrum of the conventional Fig. 9. Harmonic spectrum of the


SVM optimal SVM

5 Conclusion
a) The possibility of applying IA to obtain optimized SVM sequences has been invest
IAted in this paper. The OF has been defined with the goal of lowering most significant
harmonics.
b) The optimized sequence obtained using IA has been provided in this paper. This
sequence distributes the waveform energy more evenly throughout the spectrum, re-
sulting in less harmonic pollution.
Since the aim of this paper is to optimize the harmonic performance of the SVM
based output waveform, the associated switching losses are not included. Although this
paper focuses on two level SVM, the developed method can be extended to multilevel
schemes by devising proper formulation.

Acknowledgment
The authors wish to express their gratitude to the National Foundation of China, school
of electrical engineering of Wuhan University, and Hubei University of Education, for
support of this research effort (National Natural Science Foundation of China under
Grant 50807041, foundation of Key subject of Hubei University of Education, founda-
tion of the master-degree program of Hubei University of Education, foundation of the
key project of Hubei University of Education).

References
1. Holtz, J.: Pulsewidth modulation—A survey. IEEE Trans. Ind. Electron. 39(5), 410–420
(1992)
2. Schutten, M.: Genetic Algorithms for Control of Power Converters. In: Proceedings of 26th
Annual IEEE of Power Electronic SpecialistsConference.PESC 1995 Record, Atlanta, GA,
vol. 2, pp. 1321–1326 (1995)
3. Mehrizi-Sani, A., Filizadeh, S.: An Optimized Space Vector Modulation Sequence for Im-
proved Harmonic Performance. IEEE Trans.Ind. Electron. 56(8), 2894–2903 (2009)
616 S. Xiaofang et al.

4. Yuan, Yuan, J., Su, X., Chen, B., Tian, C.: Optimum Vector PWM Strategy for Three-Phase
Inverter Based on Immune Algorithm. Transactions of China Electrotechnical Soci-
ety 24(9), 114–119 (2009)
5. Yuan, J., Chen, B., Tian, C., Jia, J.: The research of Inverter’s control based on Immune
Algorithm. In: Proceedings of the CSEE, vol. 26(5), pp. 110–117 (2006)
6. Yuan, J., Chen, B., Jia, J.: Gentic Algorithm based approach for invertor control. Automa-
tion of Electric PowerSystems 28(24), 32–35 (2004)
7. Yuan, J., Chen, B.: Research on Optimum Control Strategy of Three-Level Single-Phase
Full-Bridge Inverter. Transactions of China Electrotechnical Society 21(3), 42–46 (2006)
8. Saeedifard, M., Nikkhajoei, H., Iravani, R.: A space vector modulated STATCOM based on
a three-level neutral point clamped converter. IEEE Trans. Power Del. 22(2), 1029–1039
(2007)
9. Saeedifard, M., Nikkhajoei, H., Iravani, R., Bakhshai, A.: A space vector modulation ap-
proach for a multimodule HVDC converter system. IEEE Trans. Power Del. 22(3),
1643–1654 (2007)
10. Somasekhar, V.T., Gopakumar, K., Baiju, M.R., Mohapatra, K.K., Umanand, L.: A multi-
level inverter system for an induction motor with open-end windings. IEEE Trans. Ind.
Electron. 52(3), 824–836 (2005)
11. Lopez, O., Alvarez, J., Doval-Gandoy, J., Freijedo, F.D.: Multilevel multiphase space vector
PWM algorithm. IEEE Trans. Ind. Electron. 55(5), 1933–1942 (2008)
12. Jiao, L., Wanlei: A novel Genetic algorithm Based on Immunity. IEEE Transactions on
System,Man,and Cybernetics-Part A:Systems and humans 30(5), 552–561 (2000)
13. Bowes, S.R., Holliday, D.: Optimal regular-sampled PWM inverter control techniques.
IEEE Trans. Ind. Electron. 54(3), 1547–1559 (2007)
14. Maswood, A.I., Wei, S.: Genetic-algorithm-based solution in PWM converter switching. In:
IEE Proceedings Electric Power Applications,, May 6, vol. 152(3), pp. 473–478 (2005)
15. Gupta, A.K., Khambadkone, A.M.: A space vector PWM scheme for multilevel inverters
based on two-level space vector PWM. IEEE Trans.Ind. Electron. 53(5), 1631–1639 (2006)
16. Lai, Y.S., Shyu, F.S.: Optimal common-mode voltage reduction PWM technique for in-
verter control with consideration of the dead-time effects—Part I: Basic development. IEEE
Trans. Ind. Appl. 40(6), 1605–1612 (2004)
17. Beig, A.R., Narayanan, G., Ranganathan, V.T.: Modified SVPWM algorithm for three level
VSI with synchronized and symmetrical waveforms. IEEE Trans. Ind. Electron. 54(1),
486–494 (2007)
18. Chen, B.-Y., Lai., Y.-S.: Switching Control technique of phase-shift- controlled full-bridge
converter to improve efficiency under light-load and stand by conditions without additional
auxiliary components. IEEE Trans. Power Electron 25(4), 1001–1012 (2009)
19. Oggier, G.G., Garcıa, G.O., Oliva, A.R.: Switching control strategy to minimize dual active
bridge converter losses. IEEE Trans. Power Electron 24(7), 1826–1837 (2009)
20. Mao, X., Ayyanar, R., Krishnamurthy, H.K.: Optimal variable switching frequency scheme
for reducing switching loss in single-phase inverters based on time-domain ripple analysis.
IEEE Trans. Power Electron 24(4), 991–1001 (2009)
21. Lai, R., Wang, F., Burgos, R., Pei, Y., Boroyevich, D.: A systematic topology evaluation
methodology for high-density Three-phase pwm AC-AC converters. IEEE Trans. Power
Electron 23(6), 2665–2680 (2008)

You might also like