I. INTRODUCTION
As DC to AC power converters feeding power to AC 11. REQUIREMENTS OF THE CONTROL SYSTEM
supply systems become more numerous, the issues relating to The control of inverters used to supply power to an AC
their control need to be addressed in greater detail. Examples system in a distributed environment should be bascd on
of such inverters feeding power to AC supply systems include information available locally at the invertcr. In typical
large, distributed Uninterruptible Power Supply (UPS) powersystems, large distances between inverters may make
systems and Photovoltaic systcms. In addition, over the past communication of information between inverters impractical.
several years, there has been considerable interest in applying Communication of information may be used to enhance
inverter technology to low voltage DC (LVDC) meshed system performance, but must not be critical for systcm
power transmission systems. The feasibility from the conlrol operation. This essentially implies that inverter control
viewpoint of an LVDC mesh has been demonstrated in [l]. should be based on terminal quantities.
The transmission system could typically consist of inverters
connected at several points on the LVDC mesh, and providing Inverter
power to AC systems which could be interconnected as well. Lf
Multiple inverters connected to a common AC system 
V T E
Stiff AC
System
essentially operate in parallel, and need to be controlled in a
manner which ensures stable operation and prevents inverter
overloads. Although inverter topologies used for power
transmission have traditionally been currcnt sourced, in
recent years Voltage Source Inverters (VSI) have been
increasingly used for high power applications like electric
traction and mill drives, photovoltaic power systems and
battery storage systems. Control schemes for VSI's in power Figure 1: Inverter Connected to Stiff AC System
system environments have formed the topic of reccnt work
[2]. Further, with inverter topologies like the NeutralPoint It is well known that stable operation of a power system
Clamped (NPC) inverter [3], it is possible to achieve needs good control of the real power flow (P) and the
substantial harmonic reduction at reasonably low Pulse Width reactive power flow (Q). The P and Q flows in an AC system
Modulation (PWM) switching frequcncics. are decoupled to a good extent [4]. P depends prcdominantly
on the power angle, and Q depends predominantly on the
A standalone AC system may be described as one in voltage magnitude. This is illustrated in Figure 2. It is
which the entire AC power is delivered to the system through essential to have good control of the power angle and the
inverters. Thus, there are no synchronous alternators present volhage level by means of the inverter. Control of frequency
in the system which would provide a refcrence for the system dynamically controls the power anglc, and thus thc real
frequency and voltage. All invertcrs in the systcm necd to be power flow. To avoid overloading the inverters, it is
operated to provide a stable frequency and voltage in the important to ensure that changes in load are takcn up by the
prcscnce of arbitrarily varying loads. This paper first inverters in a prcdctcrmincd manner, without
dcvclops a control nicthod for an invcrtcr fccding rcal and communication. This is achicvcd in conventional power
reactive power into a stiff AC system with a defined voltage, systems with multiple generators by introducing a droop in
as shown in Figure 1. This forms the basis of a control the frequency of each generator with the real power P
@7803O4535/91$1
.o@1991IEEE 1003
delivered by the generator [4]. This permits each generator to
take up changes in total load in a manner determined by its
frequency droop characteristics, and essentially utilizes the
system frequency as a communication link between the 1 2 3 4
generator control systems. The same philosophy is used in
this paper to ensure reasonable distribution of total power
bctween parallelconnected inverters in a standalone AC
system. Similarly, a droop in the voltage with reactive power
is used to ensure reactive power sharing.

V Figure 3b: Inverter Switch Positions
3
I
I 2
p==sins
0 Lf
Q= 2  YE cos6
0 Lf 0 Lf
...(4)
...(3)
In these equations, the quantity 'fgenerically denotes a
physical quantity, such as a voltage or a current. In the ...(5 )
absence of a neutral connection, the quantity 'fn ' is of no 
interest. For a sixpulse VSI, the inverter output voltage The magnitude of '+'vis:
space vector can take any of seven positions in the plane
specified by the dq coordinates. These are shown in Figure 3
as the vectors 0  6. The timeintegral of the inverter output ...(6 )
1004

PI
Regulator

Lhoica
of
V, i e.
vb eb
Vc 1 ec
FEEDBACK
Figure 4: Inverte~Con~rolScheme

The angle of WV with respect to the qaxis is:
..(11)
6, =wl (w) In this expression, we and WV are the magnitudes of the
vgv ...(7)
AC system and the inverter flux vectors respectively, and S,
The d ann q axis components of the AC system voltage flux is the spatial angle between the two flux vectors. o is the
vector, ye, its magnitude anAangle,are defined in a similar frequency of rotation of the two flux vectors. The expression
for reactive power transfer for Figure 1 can be derived in a
manner. The angle between y v and w e is defined as: similar manner. This is:
$=S, 6e ...(8)
Control of the flux vector has been shown to have ...(12)
good dynamic and steadystate performance [5. 61. It also Equations (11) and (12) indicate that P can be
provides a convenient means to define the power angle, since controlled by controlling 6, , which can be defined as the
the inverter voltage vector switches position in the dq plane,
while there is no discontinuity in the inverter flux vector. It power angle, and Q can be controlled by controlling y v . The
is useful to develop the power transfer relationships in terms crosscoupling between the control of P and Q is also
of the flux vectors. The basic real power transfer relationship apparent from these equations.
for the system of Figure 1 in the dq reference frame is: The control system for the inverter is given in Figure
4. The two variables that are controlled directly by the
...(9) inverter are '+'v and S, . The vector y v is controlled so as to
have a specified magnitude and? specified position relative to
In (9). eq and ed are the q and d axis components
respectively of the AC system voltage vector E. Also, iq and the AC system flux vector '4%. This control forms the
innermost control loop, and is very fast. It is noted that both
id are the components of the current vector i. When iq and id the inverter and the AC system voltage space vectors are
are expressed in terms of the fluxes, the equation is expressed obtained by measuring instantaneous voltage values which arc
as: available locally. The set points for the controller are P* and
Q*, and the set points for the innermost control loop, y; and
are derived from these. The actual values of P and Q
calculated from the feedback are compared with the set
Taking into account the spatial relationships bctwecn values. The error drivcs a ProportionalIntegral (PI)
the two flux vectors and assuming the AC system voltage to
be sinusoidal, (10) can be expressed as: regulator, which generates the sct points d
and 6*, for the
I

innermost control loop. The control of the invertcr to inverter flux vector YV.The locus is sccn to be close to a
generate the specified Vv and 6p is detailed in the next sub circle, since the magnitude v v very tightly controlled. Figure
section. 6 shows the inverter linetoline voltage v& and the invcrtcr
line current ia for P* = 1 MW and Q* = 500 kVAR. Figure
A. Control of vvand 6,
7 shows the response of the inverter to step changes in Q*
The control of Vvand 6p forms the first level of and P* successively. It is noted that there is a disturbance in P
when Q* is changed and a disturbance in Q when P* is
control, and directly controls the inverter switching. The
changed. In each case, the PI regulators modify the set
choice of the inverter switching vector is made on the basis of
values of 6*, and $ to maintain the P and the Q at thc set
the deviations of wv and 6p from the set values Y': and ,
values. Also, the tight control of P and Q within limits is
and the position of the inverter flux vector in the dq plane. apparent from Figure 7.
.
given by & If the deviation of 6p from 6*p is more than a
specified limit, a zero switching vector is chosen. If this
deviation is less than a specified limit, or if YVdeviates from
4'6 by more than a specified amount, a switching vector is
chosen which increases 6, and changes yrv in the correct
direction. This is essentially accomplished by hysteresis
comparators for the set values, and then using a lookup table
to choose the correct inverter output voltage vector . The
considerations for developing the lookup table are dealt with
in [5].The choice of inverter switching vector is dictated by
.
the value of 6v The dq plane is divided into six sectors for
6~ as shown in Figure 3a, which also shows the inverter
switching vectors. The inverter switch positions for the
vectors are shown in Figure 3b. The value of 6v determines
the choice of two possible inverter switching vectors, apart
from the zero vector. One vector increases the magnitude Vv
and the other decreases it, while both tend to increase 6,.
Thus, to decrease S, , the zero switching vector is chosen. To  ,
L10.O 6.00 i,oo i 00 6.00 1'0 0
correct the value of vv, one of the two active switching YCP
vectors is chosen depending on the sign of the correction
required. Table1 gives the choice of active vectors for given Figure. 5 : Invcrter RUXVector
.
positions of the inverter flux vector, specified by 6v In this
manner, Vv and 6 p are tightly controlled to lie within IV. CONTROL OF INVERTERS IN A STANDALONE
specified hysteresis bands by means of inverter switching. SYSTEM
The tip of the inverter flux vector is guided along an almost
The control of a single inverter feeding a stiff AC
circular path. Control of Yv' and S, in this manner results in a system based only on instantaneous mcasurement of terminal
PWM voltage waveform at the inverter output. quantities now forms the basis of the control scheme for
multiple inverters in standalone system environments. The
Table1: Choice of Switchine Vector
essential difference in the control scheme is that in the stand
alone system, there is no AC side voltage available for
Sector NO. (Location of WV reference. The inverters themselves produce the AC system
voltage, which is fcd back to control the inverters. There is
thus a possibility of controlling the voltage and the frequcncy
of the AC system by inverter control. Figure 8 shows two
inverters feeding into a standalone AC system. The invertcrs
are interfaced to the AC systcm through LC filters. The two
inverters are connccted by a tie line, and each invertcr has a
local load. The DC power source rcprescnts a 10 kV DC
(The zero vector is chosen to decrease Ep) power transmission mcsh. The nominal voltage on the AC
systcm is 3.6 kV rms linctoline, and the nominal frcqucncy
B . Sitnulalion Results is 60 Hz. Each inverter is a six pulse VSI made up of GTO
Simulation results of the control scheme of Figure 4 switches.
applied to the power system of Figure 1 are presented in As in the single inverter case, the two variables that are
Figures 5.6, and 7. The DC bus voltage is taken to bc 10 kV directly controlled are YVand 6p for each invertcr. Outcr
and the linetoline voltage of the AC system is taken to be
3.3 kV rms. Figure 5 gives the plot of the locus of the
0 ul
0
N
r*:
a
>
WLn
E!N
U
0
0
0
v)
I I
I
From Filter Output
Figuro 9: Frcqucncy Controller for Standhlonc Systcm figurc 10: Voltage Conirollcr for SimidhloncSystcm
0 0
0 In
' 1
Figure 11: Inverter Real and Reaaive Power (StandAlone System) I
Q o l = 0.2MVAR Q02 = 0.1 MVAR presented indicate that the scheme effectively achieves the
n i =1.0 x 104 n2 =2.0 x 104 goals of power sharing in the presence of arbitrarily
changing loads. Active damping in the loop formed by the
The nominal voltage is 3.6 kV rms linetoline and the filter capacitors and the tie line, and the use of a frcqucncy
nominal frequency is 60 Hz. The filter components for the observer would enhance the performance further. Thcse
two invcrter systems are identical. Figure 11 shows the form the goals of a later paper.
rcsponse of the inverters when the resistance RE2 (Figure 8)
is decreased suddenly to half its value. Figure 11 shows the
rcal and reactive powers supplicd by the two invericr systems VI. ACKNOWLEDGEMENTS
to the load. Figure 11 shows that Inverter 1 carries a larger
share of the real power, since it has a stiffer slope. Figure 12 This project is currcntly funded by grant #8818339 of
shows the linetoline voltage across the filter capacitor of the National Science Foundation and Agreement RP79 1112 of
Inverter 1. The plot for the reactive powers in Figure 11 the Electric Power Research Institute.
shows oscillations. These oscillations occur in the absence of
active damping of the loop formed by the two filter
capacitors and the tieline inductance. VU. REFERENCES
[ l ] B. K. Johnson, R. 11. Lasseter. R. Adapa. "Power Control
A plications on a Superconducting LVdc Mesh', 90 S M 3350 PWKD.
&E/lJES 1990 Summer Meeting. July 1990 (to appear in IEEE Trans. on
Powcx Delivery).
[2] L. Angquist, L. Lindbcrg. "Inner'Phase Angle Control of Voltage
Source Converter in High Power Applications", To be presented at the
E E E Power Electronics Specialists Confercnce, 1991.
[3] A. Nabae, I. Takahashi, 11. Akagi. "A NeutralPointClamped PWM
Inverter", IEEE Trans. Ind. Appl., vol. IA17, pp. 518523. Sept/Oct
1981.
[4] A. R. Uergcn, "Power System Analysis". PrcnticeIId, Inc., 1986.
[SI 1. Takahashi and T. Noguchi, "A New QuickResponse and Iligh
Figure 1 2 Voltage across Inverter 1 Filta Capacitor Elficiency Control Strategy of an Induction Motor", ICEE Trans. lnd.
Appl., vol. IA22, pp. 820827. Sep/Oct 1986.
[6] M. Dcpcnbrock. "Direct SelfControl (DSC) of InverterFed Induction
Machine", IECE Trans. Power Electronics, vol. 3, pp. 420429. Oct
1988.
V. CONCLUSIONS
[7] T. A. Lipo, "Analysis of Synchronous Machines", Course Notes,
This paper has dcscribcd a method to effectively Univcrsity of WisconsinMadison, 1990.
control inverters in a standalone AC supply system, without
any form of signal communication. The simulation results