You are on page 1of 8

Coordinated Control of Distributed Generation

Karina Mu noz and Kevin Wedeward


Department of Electrical Engineering
New Mexico Institute of Mining and Technology
Socorro, New Mexico 87801
Email: kmunoz@nmt.edu, wedeward@nmt.edu
AbstractThis paper presents an approach to coordinate Dis-
tributed Generation (DG) such that quantities in an unbalanced
distribution system are balanced and power factor corrected.
Models for the DG and three-phase distribution network are
combined to form a set of Differential Algebraic Equations
(DAE) that represent the behavior of the distribution system. Two
state-feedback controllers with reference inputs are presented for
coordinated control of DG. Both are designed via linearization
of the DAE model. The rst uses an additional matrix selected so
that the outputs match the reference values in steady-state while
the second uses an internal model design with integral to drive the
outputs to the reference values. The approach is demonstrated on
models of 5-bus and 10-bus test systems with inverter-connected
DG. The control objective for the 5-bus test system is to balance
active line powers across the three phases as supplied from the
transmission system. The control objective for the 10-bus system
is to correct power factor on each phase as supplied from the
transmission system. Robustness of the internal model controller
is investigated for uncertainties (up to 25%) in line parameters
and unknown step changes (up to 25%) in loads, and results
show that control objectives are achieved in the presence of these
uncertainties.
I. INTRODUCTION
As demand for power and concern about the environment
increase, Distributed Generation (DG) is becoming more at-
tractive. Installing DG at the distribution level is often less
burdensome than expanding the transmission system, which
can be difcult and expensive due to the cost of system
components and uctuating environmental regulations [1].
Additional benets associated with DG include reduction in
power demand from the transmission system, greater control
over costs and sources and less transmission losses. DG is
also associated with improved power reliability and higher
voltages, convenient for critical and sensitive loads. Finally,
DG such as photovoltaics, wind turbines and fuel cells reduce
green house emissions by producing cleaner energy.
While the commonly recognized benets associated with
DG discussed above are numerous, DG can also be utilized
to perform ancillary services to the main grid for applications
such as voltage regulation, power factor correction, load bal-
ancing and load shaping [2][5]. There have been numerous
applications researched for using DG to perform functions
other than providing power in place of the transmission
network. In [6], both active and reactive power are controlled
to enhance the stability of the system and improve the quality
of the power being delivered by compensating for any load
increase or disturbances at the local level. A parameter es-
timation and feedforward method is used to reduce transient
coupling and overshoot when tracking the reference values.
The controller is tested under grid-connected and island mode.
In [3] and [7], DG is used for voltage regulation and the idea of
balancing voltages at each phase using independent controllers
is presented in [3]. In [8], fuel cells, a common type of DG,
are used to reduce oscillations caused by incidents such as
sudden load increase. The controller model takes advantage
of the fast response of the fuel cell. Frequency control for
system stability is discussed in [9]. In [1] and [10] control of
DG connected in parallel is used for island mode systems in
order to provide an optimal load sharing technique. Finally, in
[11], DG is used for microgrids to aid in the transition between
grid-connected and islanded mode.
This paper presents an approach to coordinate DG to
balance characteristics among an unbalanced distribution sys-
tem and to correct power factor via state-feedback control.
Assuming quasi-steady state, an unbalanced three-phase dis-
tribution network will be modeled with algebraic equations,
while dynamic behavior of DG will be modeled with ordinary
differential equations. The controller takes measurements from
the network (e.g., voltages or powers) as control objectives and
uses these to adjust setpoints for all DG. Linearization of the
DAE model, including the algebraic equations for the network
and the differential equations for the DG, is used to design
two different controllers which are applied to the nonlinear
system. Robustness of one of the controllers is tested against
model uncertainties such as network parameters and outside
disturbances such as a change in load. Both controllers are
simulated on two different distribution systems and results are
presented.
II. BACKGROUND
In this section, a model for the distribution system consisting
of the three-phase network and DG is presented. Loads are
assumed to be functions of time.
A. Distribution Network Model
General equations for three-phase network devices such as
transmission lines, transformers and voltage regulators can
be found in [12]. These equations that describe three-phase,
voltage-current relationships for individual components can
be combined and converted to complex power equations that
represent power balance on every phase at every bus in
the network. The average power balance and reactive power
978-1-4244-6551-4/10/$26.00 2010 IEEE
balance equations for every phase at every bus can be written
as follows:
P
k
= |

V
k
|
3N

n=1
|

Y
kn
||

V
n
|cos(

V
k


V
n


Y
kn
) (1)
Q
k
= |

V
k
|
3N

n=1
|

Y
kn
||

V
n
|sin(

V
k


V
n


Y
kn
) (2)
where N is the total number of busses in the distribution
system; k = 1, 2, . . . , 3N, is the index used to represent every
phase at every bus; P
k
, Q
k
are the average and reactive power,
respectively, injected into a phase at a bus;

V
k
is the phasor
representation of the sinusoidal voltage on a phase at a bus;

Y
kn
is the admittance between one phase and another phase
at the same or a different bus; | | denotes the magnitude of a
complex quantity; and

denotes the phase angle of a complex
quantity.
B. Models of Inverter Connected Sources
DG such as fuel cells and photovoltaic systems generate
electricity at Direct Current (DC), thus requiring an inverter
to convert DC to to Alternating Current (AC) for connection to
the grid. Several models for inverters have been proposed [10],
[11], [13], and two representative examples are considered in
this paper and briey described below.
1) Inverter Model 1: The rst inverter model is presented in
[11]. The main objective of this inverter and its controller are
to control the active power delivered to the network as well as
the voltage at the terminal bus. The terminal bus is dened as
the single-phase, AC bus on the distribution network to which
the DG is connected. The inverter bus, where DC is converted
to AC, is connected to the terminal bus via a transformer with
series branch impedance jX . Equations that describe the
inverter, controller and network are as follows:
m = K
1
(V
set
|

V
t
|) (3)

= K
2
(P
0
R(v + K
4
) P
gen
) (4)
v = K
3
(

V
t

p
) (5)

p
= v + K
4
(6)

V
i
= +
p
(7)
|

V
i
| =
mV
dc
V
base
(8)
P
gen
=
|

V
i
||

V
t
|
X
sin(

V
i


V
t
) (9)
Q
gen
=
|

V
i
||

V
t
|
X
cos(

V
t


V
i
)
|

V
t
|
2
X
(10)
where P
gen
, Q
gen
are the real and reactive power, respectively,
injected by the generator into the network. Equations (9) and
(10) are the only tie between the generator and the distribution
network.

V
i
,

V
t
are the phasors representing the sinusoidal voltages
at the inverter and terminal buses, respectively. m is the
modulation index of the inverter and essentially establishes
the voltage magnitude at the AC side. =

V
i

p
is the
difference between the phase angle at the inverter bus and
reference phase angle
p
from a phase-locked loop. The phase-
locked loop is used to help the inverter synchronize with the
AC side. v is an intermediate variable associated with damping
in the controller, V
set
is the setpoint for the desired magnitude
of the voltage at the terminal bus, P
0
is the nominal real
power output of the generator, R is the droop constant, V
base
is the base voltage for the inverter, V
dc
is the DC voltage
on the DC side of the inverter,
p
is the estimated frequency
deviation, and K
1
, K
2
, K
3
and K
4
are gains. Equations (3)-(5)
make use of integral control in order to match up the voltage
magnitude |

V
t
|, the power P
gen
and the angle
p
with V
set
,
P
set
= P
0
R(v + K
4
) and

V
t
, respectively.
2) Inverter Model 2: The second inverter model is pre-
sented in [10]. The time-domain version is considered in this
paper with an adaptation of the governing equations as follows
=
f
(k
p
P + (
o
)) (11)
|

E| =
f
(k
v
Q + (|

E| E
o
)) (12)

E =

E
=
ref
(13)
where ,

E are the output frequency and voltage (as a phasor)
from the inverter, respectively;
f
is the cutoff frequency
of a low-pass lter; k
p
, k
v
are frequency and voltage droop
coefcients, respectively; P, Q are average and reactive power
injected by the inverter;
o
, E
o
are the frequency and voltage
setpoints, respectively; and
ref
is the reference frequency of
the transmission network. The inverter can be considered as
an ideal voltage source where amplitude, phase and frequency
can be controlled.
C. DAE Modeling
Once algebraic equations for the network and the differential
equations for the inverters have been identied, they can
be combined with load models (here functions of time) to
form a system of Differential Algebraic Equations (DAE) that
represent the complete distribution system. If equations that
represent outputs y
out
to be controlled are also included, the
following system of equations is obtained:
x = f(x, y, u
f
) (14)
0 = g(x, y, u
g
) (15)
y
out
= h(x, y) (16)
where the vector function f(x, y, u
f
) represents the differen-
tial equations of all inverters given by equations (3)-(6) for
Inverter Model 1 and by (11)-(13) for Inverter Model 2; the
vector function g(x, y, u
g
) represents the algebraic network
equations (1) and (2) for all phases and buses; vector function
h(x, y) represents the equations of outputs to be controlled
and will be seen in later examples; x are the states of the
model; y are the algebraic variables; y
out
are the outputs
to be controlled; and u
f
and u
g
are the inputs with respect
to f and g, respectively. It is important to note that quasi
steady-state is assumed for the network; therefore, although
variables such as

V ,

Y , P, Q in equations (1) and (2) vary, the
frequency is assumed to be xed. The DAE model comprised
of equations (14)-(16) will serve as the system for which
coordinated controllers are designed in the next section.
III. STATE-FEEDBACK CONTROL DESIGN
This section presents approaches to design two state-
feedback controllers. The process begins by linearizing the
nonlinear DAE model consisting of the differential equations
(14) describing the DG, the algebraic equations (15) describing
the network and the equations (16) describing the outputs to
be controlled. Once the system is linearized, a controller can
be designed for the linear system and applied to the nonlinear
system.
When there are several DG within a distribution system,
all DG aid in the process of meeting the control objective
and the term coordinated control of DG may be applied.
A conceptual diagram of coordinated control of DG within
a distribution system is shown in Figure 1. The controller
takes measurements of quantities (e.g., line powers, power
factor, bus voltages, etc.) from the distribution system. These
measurements are used by the controller to specify setpoints
(e.g., voltage and power) for the DG to meet control objectives.
Fig. 1. Conceptual diagram of coordinated control of DG
A. Initial Conditions
Linearization begins by nding an operating point about
which to linearize. The operating point is selected here by
nding a load ow solution to the distribution systems power
balance equations and then computing the corresponding
initial conditions for the DG such that the system is at
equilibrium. Initial conditions for the algebraic variables, y
o
,
come from a load ow solution, or equivalently, nding a
solution to the nonlinear equations for the network given by
(1) and (2).
1) Inverter Model 1 Initial Conditions: Initial conditions
for the algebraic variables, y
o
, can be used to nd initial
conditions for the states, x
o
, and setpoints of the inverter.
Differential equations describing Inverter Model 1 are given
by equations (3)-(6) which can be set to zero for equilibrium
and solved for the initial conditions. Here it is assumed that a
load ow solution has been computed from which the complex
power

S
gen
= P
gen
+ jQ
gen
injected by the generator and
terminal voltage

V
t
are specied. Following some algebra, the
complex generator power and terminal voltage can be used to
nd the magnitude and phase angle of the phasor representing
the voltage at the inverter bus. Equations (9) and (10) are
manipulated to nd
|

V
i
| =

X
2
_
P
2
gen
+ Q
2
gen
_
+ 2|

V
t
|
2
XQ
gen
+|

V
t
|
4
|

V
t
|
2

V
i
=

V
t
tan
1
_
P
gen
X
XQ
gen
+|

V
t
|
_
With |

V
i
|,

V
i
known, equations (3)-(6) are set to zero
and combined with (7) and (8) to nd the following initial
conditions for the states and setpoints.
m
o
=
|

V
i
|V
base
V
dc
(17)

po
=

V
t
(18)

o
=

V
i

po
=

V
i


V
t
(19)
v
o
= K
4

o
= K
4
(

V
i


V
t
) (20)
V
set
= |

V
t
| (21)
P
0
o
= R(v
o
+ K
4

o
) + P
gen
= P
gen
(22)
Initial conditions for the outputs to be controlled, y
outo
, can
be found by substituting in the initial conditions for y
o
and
x
o
into their particular equation (16).
2) Inverter Model 2 Initial Conditions: Initial conditions
for the states, x
o
, for Inverter Model 2 described by equations
(11)-(13) can also be found using y
o
given by the load
ow solution. Since the load ow solution gives the terminal
voltage (here denoted

E), initial conditions for |

E| and

E
are simply
|

E
o
| = |

E| (23)

E
o
=
E
=

E (24)
where

E is the complex voltage at the bus where the dis-
tributed generator is connected. Initial conditions for the
setpoints
o
and E
o
can be found by setting equations (11)
and (12) equal to zero and are given by:

o
=
n
+ k
p
P
gen
(25)
E
o
= |

E| + k
v
Q
gen
(26)
The initial condition for ,
o
, is simply the nominal fre-
quency
ref
.
B. Linearization
The nonlinear system given in equations (14)-(16) may be
linearized by letting x = x

+ x, y = y

+ y, y
out
=
y
out
+y
out
, u
f
= u
f
+u
f
and u
g
= u
g
+u
g
where x

,
y

, y
out
, u
f
and u
g
are the operating point and equivalently
the initial conditions found in the previous section, and x,
y, y
out
, u
f
, and u
g
are small deviations from the
operating points x

, y

, y
out
, u
f
and u
g
, respectively.
Taking the Taylor Series Expansion about the operating point
and ignoring 2
nd
and higher order terms, equations (14), (15)
and (16) become:
x =
f
x
x +
f
y
y +
f
u
f
u
f
(27)
0 =
g
x
x +
g
y
y +
g
u
g
u
g
(28)
y
out
=
h
x
x +
h
y
y (29)
where the notation
a
b
is the partial derivative of the vector
function a with respect to the vector of variables b evaluated
at the operating point. Solving for y in equation (28) and
substituting into equations (27) and (29) yields a linearized
mathematical model of the system in the general state-space
form
x = Ax + B
1
u
f
+ B
2
u
g
(30)
y
out
= Cx + Du
g
(31)
where A =
_
f
x

f
y
_
g
y
_
1
g
x
_
, B
1
=
f
u
f
, B
2
=

_
f
y
_
g
y
_
1
g
ug
_
, C =
_
h
x

h
y
_
g
y
_
1
g
x
_
and
D =
_
h
y
_
g
y
_
1
g
ug
_
all of which are constant matrices
that result from evaluation at the operating point.
C. Controller Design
Two state-feedback controllers, both based upon the lin-
earized system above are presented below.
1) State-Feedback Controller with Reference Input: Using
the linearized equations (30) and (31) a state-feedback con-
troller with reference input can be dened as:
u
f
= Kx +

Nr (32)
where K is a matrix of gains with appropriate dimensions used
to select the closed-loop poles of the system and

N is a scaling
matrix used to set constant y
out
equal to constant reference
input r in steady-state. Substituting the controller dened by
equation (32) into the system dened by equations (30) and
(31) yields the closed-loop system:
x = (AB
1
K)x + B
1

Nr + B
2
u
g
(33)
y
out
= Cx + Du
g
(34)
To nd

N, steady-state is assumed such that x = 0 in
(33) and y
out
= r in (34). Assuming that u
g
is known
and is equal to some constant l times r,

N can be found to
be

N =
_
C(AB
1
K)
1
B
1
_
1

_
I + l
_
C(AB
1
K)
1
B
2
D
__
(35)
It is important to note that

N can only be obtained if the
disturbance u
g
and system matrices are known. Therefore,
the state-feedback controller with reference input will not work
for a system with unknown disturbances or uncertainty in
parameters.
2) Internal Model Controller: Although the state-feedback
controller with reference input is simple, many times a dif-
ferent controller is more appropriate, especially when the
input u
g
is unknown or there is uncertainty in the model.
An internal model controller is developed below to provide
tracking capabilities and robustness to uncertainties related to
the model and/or disturbances.
Let the tracking error, e, be dened as y
out
r, where
y
out
is the deviation in output and r is the deviation in
reference value. For a step reference value, the derivative of
the tracking error becomes:
e = y
out
= C x + D u
g
(36)
Equation (36) can be combined with the derivative of equation
(30) to form a new augmented system also based upon the
tracking error. The new system can be described by the
following two equations:
e = C x + D u
g
(37)
x = A x + B
1
u
f
+ B
2
u
g
(38)
Assigning z = x, the system above becomes:
e = Cz + D u
g
(39)
z = Az + B
1
u
f
+ B
2
u
g
(40)
which can be written in the general form of equation (30) as:
_
e
z
_
=
_
0 C
0 A
_ _
e
z
_
+
_
0
B
1
_
u
f
+
_
0
B
2
_
u
g
(41)
This new system can be used to design a controller with
control law:
u
f
= [K
e
, K
z
]
_
e
z
_
= K
e
e K
z
z (42)
or equivalently (via integration)
u
f
= K
e
_
edt K
z
x (43)
where K
e
and K
z
are gain matrices of appropriate size. If
this new augmented system is stable and u
g
is constant or
slowly varying, e 0 as will z = x. For a higher order
reference input, additional integral terms will be necessary
[14].
IV. EXAMPLES WITH SIMULATION RESULTS
This section presents simulation and analysis of two dis-
tribution systems (a 5-bus and a 10-bus) that implement the
approaches to control described in Section III. For simplicity,
only one type of DG is used in each distribution system.
Control objectives are dened for each system and the design
process summarized.
A. 5-Bus Distribution System
Fig. 2. Diagram of 5-bus distribution system
The rst distribution system is shown in Figure 2. The
system consists of ve buses, three loads and three distributed
generators. It is assumed that the distribution system is directly
connected to the transmission system, and therefore bus num-
ber one is modeled as an innite bus. Loads and DG have
been placed at buses three, four and ve. All three of the
DG were of type Inverter Model 1 [11], and can be described
by equations (3)-(10). The system has unbalanced loads and
unbalanced impedances among phases. Parameters for all three
distributed generators are given in Table I. Network parameters
TABLE I
DG PARAMETERS FOR 5-BUS DISTRIBUTION SYSTEM
K1 10
K2 20
K3 20
K4 10
R 0.4 p.u.
X 0.2 p.u.
Vset 0.96 p.u.
V
dc
480 V
V
base
240 V
for the distribution network are given in Table II with all values
in p.u.
TABLE II
NETWORK PARAMETERS FOR 5-BUS DISTRIBUTION SYSTEM
S
Load1
0.5 + 0.25i
S
Load2
0.6 + 0.25i
S
Load3
0.55 + 0.2i
Z
23
0.0909 + 0.1818i
Z
24
0.1000 + 0.2000i
Z
25
0.1111 + 0.2222i
Z
12
_
0.22 + 0.31i 0.25 + 0.02i 0.25 + 0.002i
0.25 + 0.02i 0.2 + 0.3i 0.0025 + 0.002i
0.25 + 0.02i 0.0025 + 0.002i 0.2 + 0.28i
_
The controller used was the state-feedback controller with
reference input given by equation (32). The control objective
was to balance the active line powers as seen from the
transmission system to a desired value of 0.3 p.u. for all three
phases. A load ow was performed to nd initial conditions for
the algebraic variables, y
o
. These were then used to solve for
the initial conditions for the states of the differential equations,
x
o
, using equations (17)-(22).
The nonlinear equations for outputs to be controller, y
out
,
were dened as those needed to compute the active line powers
P
Line1a
, P
Line
1b
, and P
Line1c
provided by the transmission
system (innite bus). System variables for the DAE model
given in general form in equations (14)-(16) are as follows:
x = [m
1
,
1
, v
1
,
p1
, m
2
,
2
, v
2
,
p2
, m
3
,
3
, v
3
,
p3
]
y = [

V
2a
, |

V
2a
|,

V
2b
, |

V
2b
|,

V
2c
, |

V
2c
|,

V
3
, |

V
3
|,

V
4
, |

V
4
|,

V
5
, |

V
5
|]
y
out
= [P
Line1a
, P
Line
1b
, P
Line1c
]
u
f
= [P
0
1
, P
0
2
, P
0
3
]
The assumption that there were no inputs with respect to g
(u
g
= 0) was made and equations (30) and (31) for the
linearized system were reduced to:
x = Ax + B
1
u
f
(44)
y
out
= Cx (45)
Once the controller was designed with a response time similar
to that of the uncontrolled system in mind, it was applied to
the nonlinear system. Simulation results are shown in Figure
3 where it is noted that the controller was turned on at 5
seconds. Figure 3 shows the active line powers for all three
Fig. 3. Line powers delivered from the innite bus for 5-bus distribution
system using state-feedback with reference input controller
phases as seen from the transmission system (between buses
one and two). Although initially they are quite different, at
about 8 seconds, 3 seconds after the controller was started,
the line powers are extremely close to the desired value of 0.3
p.u. for all three phases.
B. 10-Bus Distribution System
The second distribution system studied is shown in Figure 4
and was obtained from [15]. The distribution system consists
of ten buses, nine loads and three distributed generators. The
Fig. 4. 10-bus distribution system
DG are of type Inverter Model 2 described by equations (11)-
(13). Parameters for the DG can be found in Table III and
network data are given in Tables IV and V.
TABLE III
DER PARAMETERS FOR 10-BUS DISTRIBUTION SYSTEM

ref
377 rad
kp .01
kv .01

f
37.7 rad
TABLE IV
LOADS (KW AND KVAR) FOR THE 10-BUS DISTRIBUTION SYSTEM
Phase A Phase A Phase B Phase B Phase C Phase C
Node P Q P Q P Q
2 22.5 11.25 50 25 50 12.5
3 22.5 11.25 37.5 12.5 0 0
4 22.5 11.25 25 25 25 25
5 0 0 37.5 12.5 50 12.5
6 70 58.75 0 0 0 0
7 45 33.75 0 0 25 12.5
8 45 33.75 25 12.5 0 0
9 0 0 0 0 50 37.5
10 0 0 50 37.5 0 0
TABLE V
BRANCH DATA FOR THE 10-BUS DISTRIBUTION SYSTEM IN P.U.
Branch Z
self
Z
mutual
1-2 1 + 2i 0.5i
2-3 1 + 1i 0.25i
2-4 1 + 2i 0.5i
2-5 1 + 1i 0.25i
3-6 4 + 2.25i 0
4-7 1 + 1i 0.25i
4-8 1 + 1i 0.25i
7-9 5 + 5i 0
8-10 6 + 4.5i 0
The controller used for this system was the internal model
controller described by equation (43). A load ow was per-
formed to provide initial conditions for algebraic variables y
o
from which the initial conditions for the states x
o
and setpoints
were found through equations (23)-(26). y
o
and x
o
were used
to nd initial conditions for y
out
variables taken here to be
reactive power delivered by the transmission system.
1) Power Factor Correction: The control objective for the
10-bus system was power factor correction to be achieved by
reactive power control. The nonlinear equations for the outputs
y
out
were those used to compute reactive powers Q
Line1a
,
Q
Line
1b
, and Q
Line1c
delivered by the transmission system
(innite buses). System variables for the DAE model are as
follows:
x = [
1
, |

E
1
|,

E
1
,
2
, |

E
2
|,

E
2
,
3
, |

E
3
|,

E
3
]
y = [

V
2a
, |

V
2a
|,

V
2b
, |

V
2b
|,

V
2c
, |

V
2c
|,

V
3a
, |

V
3a
|,

V
3b
, |

V
3b
|,

V
4a
, |

V
4a
|,

V
4b
, |

V
4b
|,

V
4c
, |

V
4c
|,

V
5b
, |

V
5b
|,

V
5c
, |

V
5c
|,

V
7a
, |

V
7a
|,

V
7c
, |

V
7c
|,

V
8b
, |

V
8b
|,

V
8c
, |

V
8c
|]
y
out
= [Q
Line1a
, Q
Line
1b
, Q
Line1c
]
u = [E
o
1
, E
o
2
, E
o
3
]
Again the assumption that there were no inputs with respect
to g was made and the general equations (30) and (31) for the
linearized system were reduced to:
x = Ax + Bu
f
(46)
y
out
= Cx (47)
The linear control system (43), (46), (47) was used nd gains
such that responses similar to that of the original, uncontrolled
system were achieved. The controller was then applied to the
nonlinear system at 10 seconds, and simulation results are
shown in Figure 5. Reactive power for all three phases, shown
in Figure 5, has gone down to the desired value of 0 p.u. in
about 50 seconds. This corresponds to dramatic improvement
in power factor on all three phases that become extremely
close to the desired value of one.
Fig. 5. Reactive line powers delivered from the innite bus for 10-bus
distribution system with power factor correction and internal model controller
2) Step Change in Loads: It was shown above that the in-
ternal model controller works even when there is an unknown
step change in load, or equivalently there exists an input with
respect to g, which was denoted u
g
. Because the loads are
not necessarily known, they are considered here as unknown
inputs with respect to the algebraic equations, and the linear
system used for design of the internal model controller (43)
has the form
x = Ax + B
1
u
f
+ B
2
u
g
(48)
y
out
= Cx + Du
g
(49)
To simulate a step change in load both active and reactive
power for the loads at the buses with DG were increased
by 25% at 60 seconds. Figure 6 shows reactive power as
transferred from the transmission line in the case of the
unknown step change in load. At 60 seconds, the reactive
power immediately responds to the load increase, but still
achieves the ultimate desired value of zero.
Fig. 6. Reactive line powers delivered from the innite bus for 10-bus
distribution system with 25 percent increase in load at 60 seconds
3) Uncertainty in Parameters: The internal model con-
troller was also tested against uncertainty in network pa-
rameters. Assuming all DG are tested before installation,
only network parameter uncertainty will be considered. The
linearized model with errors in the network parameters was
used to design the internal model controller for the nonlinear
system. Assuming there are no inputs with respect to g (i.e.,
u
g
= 0), the linear system used for design of the controller
will have the form
x = (A + A)x + B
1
u
f
(50)
y
out
= (C + C)x (51)
where the deviation matrices A, C are the due to the error
in the network parameters. Because B
1
is not dependent on the
algebraic variables, y, it is unaffected by changes in network
parameters. Although eigenvalues for the closed-loop system
(A B
1
K) vs. ((A + A) B
1
K) will vary, the system
given by equations (50) and (51) can still be controlled using
the internal model controller given by equation (43).
To test the effect of uncertainty in network parameters they
were randomly changed by up to 25% of their actual value
taken from a uniform distribution. Several cases were ran and
results overlaid as shown in Figure 7. Figure 7 shows the
Reactive Power with up to 25% uncertainty in the network
parameters. The control objective of reducing reactive powers
to zero for all phases is met even though the uncertainty is
present. This is due to the fact that the actual system states
and outputs are fed back to the controller and the integral term
eventually helps meet the control objective.
Fig. 7. Reactive line powers delivered from the innite bus for 10-bus
distribution system with up to 25 percent uncertainty in network parameters
with internal model controller
V. CONCLUSION
This paper examined the use of coordinated control of DG
for balancing characteristics among phases and for power
factor correction. Two state-feedback controllers with refer-
ence inputs were presented. Both controllers were designed
based upon linearization of the DAE model that consisted of
dynamic equations for the DG and algebraic equations for the
three-phase network. The rst controller used an additional
matrix selected so that the outputs match the reference in
steady-state. The second controller used an internal model
design with integral to drive the outputs to the references.
5-bus and 10-bus distribution systems were simulated and
analyzed to demonstrate the effectiveness of the controllers.
Active line powers were controlled and balanced on the 5-
bus distribution system using the state-feedback with reference
input controller. Power factor correction was performed on the
10-bus distribution system using the internal model controller
which performed well even in the presence of uncertainty in
network parameters and loads.
ACKNOWLEDGMENT
The research described in this paper was supported through
a contract with the Department of Energy.
REFERENCES
[1] J.-W. J. Mohammad N. Marwali and A. Keyhani, Control of distributed
generation systems-part II: Load sharing control, IEEE Transactions on
Power Electronics, vol. 19, no. 6, pp. 15511561, November 2004.
[2] R. H. Lasseter and P. Piagi, Extended microgrid using (DER) dis-
tributed energy resources, in 2007 IEEE Power Engineering Society
General Meeting, PES, June 2007.
[3] Y. Xu, F. Li, J. D. Kueck, and D. Tom Rizy, Experiment and sim-
ulation of dynamic voltage regulation with multiple distributed energy
resources, in 2007 iREP Symposium- Bulk Power System Dynamics and
Control - VII, Revitalizing Operational Reliability, August 2007.
[4] J. C. et. all, Ancillary services provided from DER, Oak Ridge
National Laboratory, Tech. Rep., 2005.
[5] S. R. Oliver Gehrke and P. Venne, Distributed energy resources
and control: A power system point of view, Internal Report Ris-R-
1608(EN).
[6] J.-W. J. Min Dai, Mohammad Nanda Marwali and A. Keyhani, Power
ow control of a single DG unit, IEEE Transactions on Power Elec-
tronics, vol. 23, no. 1, pp. 343351, January 2008.
[7] M. A. Kashem and G. Ledwich, Distributed generation as voltage
support for single wire earth return systems, IEEE Transactions on
Power Delivery, vol. 19, no. 3, pp. 10021011, July 2004.
[8] K. Ro and S. Rahman, Control of grid-connected fuel cell plants for
enhancement of power system stability, Renewable Energy, pp. 397
407, 2003.
[9] Y.-J. S. Xiangjun Li and S.-B. Han, Frequency control in micro-grid
power system cobined with electrolyzer system and fuzzy pi controller,
Journal of Power Sources, vol. 180, pp. 468475, February 2008.
[10] P. C. C. Ernane Antonio Alves Coelho and P. F. D. Garcia, Small-
signal stability for parallel-connected inverters in stand-alone AC supply
systems, IEEE Transactions on Industry Applications, vol. 38, no. 2,
pp. 533542, March 2002.
[11] I. A. Hiskens and E. M. Fleming, Control of inverter-connected sources
in autonomous microgrids, in Proceedings of the American Control
Conference, Seattle, WA, United States, June 2008.
[12] W. H. Kersting, Distribution System Modeling and Analysis, 2nd ed.
CRC Press, 2006.
[13] T. Green and M. Prodanovic, Control of inverter-based micro-grids,
Electric Power Systems Research, vol. 77, no. 2, pp. 12041213,
Sepetember 2007.
[14] R. C. Dorf and R. H. Bishop, Modern Control Systems, 9th ed. Prentice
Hall, 2001.
[15] A. G. E. Esther Romero Ramos and G. A. Cordero, Quasi-coupled
three-phase radial load ow, IEEE Transactions on Power Systems, June
2004.

You might also like