See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281826536

Improved Optimal Control Technique for Control of Parallel Three-phase

Inverters

Conference Paper · October 2009

CITATION READS

1 773

2 authors:

Hossein Mokhtari Hamidreza Nazaripouya

Sharif University of Technology University of California, Riverside
157 PUBLICATIONS 5,089 CITATIONS 32 PUBLICATIONS 700 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Hamidreza Nazaripouya on 16 September 2015.

The user has requested enhancement of the downloaded file.

 Improved Optimal Control Technique for Control of
Parallel Three- phase Inverters
Hamid Reza Nazari Pouya, Hossein Mokhtari, Member, IEEE
The Center of Excellence in Power System Management & Control
Sharif University of Technology
[email protected], [email protected]

Abstract- This paper proposes a high performance voltage forced to be equal with each other, thereby enhancing the
tracking and current sharing among parallel connected inverters system performance.
by applying optimal control and minimizing the cost function. For this purpose, first the three-phase system is transformed
The control system forces the voltage of load to track the voltage to a synchronous-stationary frame (dq0) to reduce the system
reference, and the current of all inverters becomes equal. to two single-phase systems. Then, the modified performance
Therefore, both current sharing and voltage tracking are
index according to state space equation of each single-phase
obtained. In addition, the smallest input energy and simplicity in
control circuit are the other advantages of the suggested method.
system and optimal methodology is defined and solved.
To show the performance of the proposed control scheme, the The suggested method offers the following advantages:
simulations with two-modular practical systems are performed
and the results are provided. The results indicate that the control 1) It improves the current sharing among the parallel
objectives are achieved even if there are parameters mismatch inverters and reduces the circulating current among
among the parallel modules. them.
2) It decreases the Total Harmonic Distortion (THD) of
the output voltage and current.
I. INTRODUCTION 3) It provides the possibility of parallel connection of
several inverters with different capacities.
The use of uninterruptable power supplies (UPSs) in order to
provide reliable and clean power to critical loads has been
increased. Therefore, as the critical loads grow, small UPS II. STRATEGY OF DESIGNING THE CONTROLLER
unit must be replaced with high capacity ones. But using high In this part, a three-phase inverter is transformed into two
power UPS systems has problems such as high initial cost, site
single-phase inverters, Then the state space equation of each
installation difficulties due to bigger size and weight and
system which includes n-single-phase parallel inverters, is
lower system reliability. On the contrary, parallel operation of
introduced. By using optimal control law, the controller is then
multi-modular UPS systems has advantages of lower cost,
easier maintenance, flexibility to increase power capability, designed.
availability and enhanced reliability. However, there is a A. Transforming a 3-phase inverter to single-phase inverters
challenging issue in designing control system of the parallel Consider the circuit topology of a 3-phase inverter
UPSs because the output voltage of all UPSs must stay connected through a 3-phase LC filter to a balance load
synchronized. Therefore, they must have the same amplitude, depicted in Fig. 1.
frequency and phase that cause equal sharing of active and
reactive power. Otherwise, the output current will contain
circulating current which may damage the power
semiconductors in parallel modules. In practice, physical
mismatches between the parallel modules and line impedances
prevent perfect load sharing. Consequently, the controller
must be designed properly such that it can provide voltage
regulation and current sharing in spite of the fact that the
inverters do not have the same parameters [1]-[7].
Applying optimal control to parallel operation of multi-
modular three-phase inverters has been studied in [8]. In this
paper, this idea is extended to parallel inverters because the Fig. 1: Three-leg 3-phase inverter circuit topology.
proposed technique in [8] cannot provide proper current
sharing when the inverter modules do not have the same The sum of three output phase voltages and the sum of three
parameters. But in the proposed approach, by adding a new output currents can be considered zero because the load is
term in the performance index, the current of inverters are balanced. Consequently, if vab and vbc are known, vca can be
calculated. Therefore, there are only two independent output Fig. 2, its eight corners, which represent 8 different states of
voltages [9]. It can be concluded that a 3-leg inverter just the voltage-vector, can be created by different switching-states
produces two independent output voltages, and if the load and [10],[11].
filter are connected in Y shape and they are balanced, then:

0
0 (1)

Then,

(2)

From (1) and (2):

(3)
Fig. 3: Switching-states/voltage-vectors in dq-plane[11].
0
If the orthogonal axes that represent a 2-dimensional plane
According to Fig. 1 and (3), phase voltages can be obtained are considered as Vq and Vd, each vector can be disintegrate
as: into two main components of Vq and Vd. Therefore, it is
2 1 1 possible to write Vq and Vd in the basis of the phase-voltage
1 2 1 (4) (van vbn vcn) by using the following transformation matrix:
1 1 2
1
If (Vag Vbg Vcg) are considered as three leg-voltages, they can √ √
be shown as three orthogonal axes that represent a 3- 0 (5)
dimensional space. Therefore, there are three independent
√ √ √
voltages that produce three phase-voltages (van vbn vcn). As it
was mentioned earlier in fact, they represent only two
independent voltages in output-voltage plane, which can be This equation is often known as the dq0 transformation [10].
projected onto a 2-dimensional plane which is orthogonal to It should be noted that V0 axis is added only to have an
the vector [1 1 1]t in the leg-voltage space. As shown in Fig. 2, inversable 3×3 square transformer matrix. This zero-sequence
Vi =100Vdc is used for convenient scaling [10]. has no effect on the system design.
Combining (4) and (5), the complete transformation from the
leg-voltage space (van vbn vcn) to dq-plane (Vq Vd) is obtained
as:

0
√ √ (6)
0 0 0

It can be seen that the first two rows of the transformation

matrix represent the three-phase voltages, i.e. va, vb, and
vc.
As a result, the leg-voltage vectors can be constructed from
the desired [Vq* Vd* V0*]T values by using the inverse of the
Fig. 2: A leg –voltage space projected onto output- voltage space [10].
Tqd0 matrix as :
1 0
According to Fig. 2 when the leg–voltage space is projected √
√
onto the output-voltage space, the balance three-phase vectors (7)
√
that stand in a 2-dimensional plane are obtained. As it is √
depicted in Fig. 3, through the mapping of the cube shown in √
 Consequently, according to (6), the desired output voltage Assuming that load is resistive, i.e. RL, the state-space
can be represented by [Vq Vd]T which is produced by equation of a multi-inverter system with n-single-phase
switching-state/voltage-vector. In fact, Tqd0 transform inverters is:
decouples the three-phase voltages, and transforms it into their
equivalent 2-dimensional constituents. Finally, the two 0 0 0 … 0 1 1 0
independent single-phase circuits instead of a 3-leg inverter 0 0 … 0 0 0
are obtained as shown in Fig. 4. Therefore, this transformation
0 0 … 0 0 0
matrix helps us to control the 3-leg inverter by controlling just
two output axes (Vq Vd ) independently. In other words, the
three-phase inverter is reduced to two independent single- 0 0 0 … 0 0
phase inverters[10].
0 … 0 0
0 0 0 … 0 0 0 1
0 0 0 … 0 0 0
0 0 … 0
0 … 0

0 … 0

Fig. 4: Equivalent dynamic model of a 3-leg inverter[10]. (8)

0 0 …
The two equivalent circuits have the components which are 0 0 … 0
equal to the original 3-leg inverter components, i.e. Ld=Lq = L, 0 0 … 0
Cd = Cq = C. 0 0 … 0

B. Optimal controller for single-phase inverters where CL= Cf1 + Cf2+ ….. +Cfn. ilj is the current of the jth
Fig. 5 depicts n-single-phase inverter models which are inverter, and ev represents the integral of the output error i.e.:
connected in parallel [12]. Mj is the gain of the SPWM and dj
is the duty cycle of the switch of the jth inverter. Lfj, Rfj, and Cfj (9)
are the inductance, resistance, and capacitance of the output
LC filter of the jth inverter respectively. and vr = Vmsinωrt is the voltage reference of all inverters.
The objective is to track a sine wave reference, and
therefore, it is clear that the issue is a tracking issue. But, the
methodology of optimal control stands on the basis of
regulation problem. Therefore, this problem must be changed
to a regulation problem first. Hence, and are added as
extra and auxiliary parameters in the state space equation to
define the voltage reference precisely.
The state-space model of an n-order system such as n-
parallel multi-inverter system can be written as:

(10)

To design the feedback control loop by using optimal

control, a performance index must be minimized [13]. In this
paper, the following performance index is proposed:

∑ (11)

where , , are real

symmetric positive semi-definite matrices, and is a
Fig. 5: Model of n-parallel inverters. real symmetric and positive definite matrix, and all the
weighting matrices are assigned by the designer.
 The performance index includes: To simplify (17), three assumptions are made. Since the term
is added as an auxiliary parameter and its coefficient
1. The input of the system whose minimization results becomes very small after optimization, its effect can be
in minimum energy and cost. neglected. It can be further assumed that from the position of
2. The weighting matrix , Adjusting its elements, the jth inverter, the feedback gains (coefficients) of other
the relative importance of the deviation of each of inverter inductor currents are the same (Filo). Finally, it is
the states from their desired values can be known that the current in the filter capacitors are very small
weighted. [13]. Therefore, the control law can be simplified as:
3. The output voltage error whose minimization,
results in desired voltage regulation.
4. The sum of current differences between the
inverters whose minimization causes desirable
current sharing among the inverters.

As compared to previous approaches, in the proposed

approach, there is a control of current error. Therefore, both
voltage regulation and current sharing performances are (18)
improved.
Since have been included in the state variables, by
defining: According to (18), just one common signal which contains
the load current information is required. Therefore, the
∑ simplicity and reliability of the control system are the other
advantages of this method.

III. SIMULATION RESULTS

0 0 … 0 1 1 0 (12)
For this part, a multi-modular system consisting of two
The performance index is simplified as: inverters is considered. The two modules are first assumed to
have the same capacity. Later, their capacities are supposed
(13) different. The system is simulated with both the proposed
method and similar method without considering the current
The advantage of this method is that the control law is a difference in the optimization function. In addition, in all
linear time invariant function of state variables [14], [15]: simulations, load changes from open circuit (no load) to a load
of R= 50Ω at t = 0.1s. System parameters are listed in Table I.

(14) TABLE I: Circuit parameter list

Parameters Values
In fact, (14) is the control law used to control the system
where K is obtained by solving Riccati equation as follows: Output filter inductor (Lf1,2) 15mH, 20mH

Resistance of the filter inductor (Rf1,2) 0.1Ω, 0.2Ω

0 (15)
Output filter capacitor (Cf1,2) 40µF
This equation can be solved by using software tools like
MATLAB. Finally, (14) can be rewritten as: Input dc voltage (Vdc) 350V

… (16) Rated load in each phase 50Ω

Substituting u = [d1 d2 …. dn]T into (16) yields: Rated output line voltage 190Vrms

Output voltage frequency 60Hz

∑ (17)
Switching frequency of the inverter (fs) 40kHz
where Fev ,Fvr ,Filj (j=1,…,n) ,Fvo,Fr are components of F.
 Simulation results of voltage and current load of the
proposed approach are shown in Fig. 6. The current of the first
inverter and the second inverter are depicted in Fig. 7. To
compare the new method with the previous one, the results of
the existing method are shown in Fig. 8.Finally, Fig. 9 shows
the difference between the current of the inverters by using the
proposed method. Fig. 10 shows the difference between the
current of inverters using the old method for the same
operating condition.
It can be seen that by applying the new method, circulating
current is 3 times less than that obtained by the old method.
Besides, the current and voltage THD are also decreased.
The results of parallel connection of two inverters with
different capacities are shown in Figs. 11, 12, 13. The first
inverter capacity is twice as big as the second ones. Fig. 8: Simulation results of two 3-phase parallel inverters with old method
when load changes from open circuit (no load) to a load of R= 50Ω. Top: the
current of first inverter. Bottom: the current of second inverter

Fig. 6: Simulation results of two 3-phase parallel inverters with new method
when load changes from open circuit (no load) to a load of R= 50Ω. Top: the
output load current. Bottom: the output voltage Fig. 9: the circulating current result in proposed method when load changes
from open circuit (no load) to a load of R= 50Ω

Fig. 7: Simulation results of two 3-phase parallel inverters with new method
when load changes from open circuit (no load) to a load of R= 50Ω. Top: the Fig. 10: the circulating current result in old method when load changes from
current of first inverter. Bottom: the current of second inverter open circuit (no load) to a load of R= 50Ω
 IV. CONCLUSION
In this paper, an optimal controller is proposed for parallel 3-
phase inverters. In this method, the feedback gain is obtained
by optimization of an improved performance index which
includes both voltage regulation and equal current sharing.
This method can further reduce the circulating current as
compared to the existing methods. The method can also
provide the control of parallel connection of inverters with
different capacities with low circulating current. In addition,
Total Harmonic Distortion of voltage and current is reduced,
and the transient respond is also improved.

REFERENCES
[1] M.C. Chandorkar, D.M.Divan and R. Adapa, “Control of parallel
Fig. 11: Simulation results of two 3-phase parallel inverters with different connected inverters in standalone ac supply systems”,IEEE Trans. Ind.
capacity (the first one is twice as big as second one) when load changes from Applicat, vol. 29, No. 1, Jan./Feb., 1993, pp. 136-143.
[2] T.F. Wu, Y.-K. Chen and Y.-H. Huang, “3C strategy for inventers in
open circuit (no load) to a load of R= 50Ω. Top: the current of first inverter.
parallel operation achieving an equal current distribution,” IEEE Trans.
Bottom: the current of second inverter Ind. Electron., vol.47, pp.273-281, Apr.2000.
[3] D. Shanxu, M. Yu, X.Jain, K. Yong, and C. Jain, “Parallel operation
control technique of voltage source inverters in UPS,” in proc. IEEE
PEDS’99, Hong Kong, pp.883-887, July 1999.
[4] T. Kawabata and S.Higashino. “Parallel operation of voltage source
inverters,” IEEE Trans. Ind. Applicant., vol.24, no.2, pp.281-287,
Mar./Apr.1988.
[5] A. Tuladhar, H. Jin, T. Unger, and K. Mauch, "Control of parallel
inverters in distributed AC power systems with consideration of line
impedance, "IEEE Trans. Ind. Applicant, vol. 36, pp. 131-138.
Jan./Feb.2000.
[6] X.Sun, Y.-S. Lee, and D. Xu, “Modeling, analysis, and implementation
of parallel multi-inverter system with instantaneous average-current-
sharing scheme,” IEEE Trans. Power Electron., vol.18, pp.844-856, May
2003.
[7] L. Chen, X. Xiao, C. Gong, and Y. Yan, “Circulating current
characteristics analysis and control strategy of parallel system based on
double close loop controlled VSI, “ in Proc. IEEE PESC’04, pp.4791-
4797, june 2004.
[8] H. R. Nazaripouya, and H. Mokhtari “Control of Parallel Three- phase
Fig. 12: Simulation results of two 3-phase parallel inverters with different Inverters Using Optimal Control and SVPWM Technique” in Proc.
capacity (the first one is twice as big as second one) when load changes from IEEE ISIE’09, pp.1823-1828, July 2009
[9] N.Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics,
open circuit (no load) to a load of R= 50Ω. Top: the output load current. Converters, Applications, and Design, 2nd ed. New York: Wiley, 1995.
Bottom: the output voltage [10] M. J. Ryan, R. D. Lorenz, and R. W. De Doncker, “Modeling of sinwave
inverters: A geometric approach,” IEEE Trans. Ind. Electron., vol. 46,
pp.1183-1191, Dec.1999
[11] Novotny, D.W., Lipo, T.A., Vector control and Dynamics of AC Drives,
Oxford University Press, Oxford, 1997.
[12] Kai Zhang, Yong Kang, Jian Xiong, and Jian Chen,”Direct repetitive
control of SPWM inverter for UPS purpose,” IEEE Trans. on Power
Electr., vol. 18, No. 3, 2003.
[13] X. Sun, L.K. Wong, Y. S. Lee, and D. Xu, “Design and analysis of an
optimal controller for parallel multi –inverter system,” IEEE
Transactions on Circuits and Systems- II, vol. 52, no.1, January 2006.
[14] B. D. O. Anderson and J. B. Moore, Linear Optimal Control. Engle-
wood Cliffs, N.J.: Prentice-Hall, 1971.
[15] D. E. Kirk. Optimal Control Theory. Prentice-Hall Inc., 1970.

Fig. 13: the difference between two inverter currents with different capacity
when load changes from open circuit (no load) to a load of R= 50Ω.

View publication stats