DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
1
Task 1: Coupling Low-Voltage Microgrids into
Mid-Voltage Distribution Systems
Zhao Wang, Student Member, IEEE, and Michael Lemmon, Member, IEEE
Abstract
This report presents results of the first two stages of the project, i.e. algorithm development and simulation
model development. A hierarchical control architecture is proposed to maximize the real power export to mid-voltage
(MV) distribution network by coupling low-voltage (LV) microgrids. At the same time, reactive power is dispatched
coordinatively so that voltage regulatory constraints are satisfied. Two levels of optimization problems are solved
to designate set points for each microsource controller: 1) the state at the point of common coupling (PCC) is
determined by each microgrid controller in a decentralized fashion, 2) set points for microsources within a microgrid
are calculated in the microgrid controller in a centralized way. Corresponding simulation model is developed in
TM
R Simulink
R with SimPowerSystems
MATLAB
toolbox. Simulation results show that the algorithm developed
can maintain system stability, while simultaneously achieving the goal of maximizing real power export under voltage
regulation constraints. Response of the system under disturbances proves the robustness of the hierarchical control
system.
Index Terms
Distributed optimization, capacity estimation, weak network.
I. I NTRODUCTION
The power grid system is expected to see tremendous changes, one of which is a high level penetration of
distributed energy resources (DERs) in LV distribution networks. DER is promoted as a means to meet the constantly
increasing demand for high quality electric power, but it may cause as many problems as it solves [1]. Current power
grid operates based on the unidirectional power flow assumption. Both control methods and protection mechanisms
are designed to work with this assumption. A high penetration of DERs in distribution networks may result in power
flow that move in the opposite direction to conventional flow situations. A “voltage rise problem” arises from the
multi-directional power flows, when integrated DERs export real power back to the distribution network [2]. Because
LV and MV segments of the power grid are weak, this problem is much more severe in these circumstances than
in the high-voltage (HV) transmission network. Since our objective is coupling LV microgrids to MV distribution
networks, we need to formulate and solve the voltage regulation problem assuming weak network connections.
Z. Wang and M. Lemmon are with the Department of Electrical Engineering, University of Notre Dame, Notre Dame, IN, 46556, USA
e-mail: wang.198@nd.edu.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
2
We intend to maximize the real power export back to the MV distribution network from the coupled LV microgrids,
while respecting regulatory constraints on voltage magnitude and frequency, transient stability in the face of faults,
topology variations, or set point changes, as well as compatibility with legacy control devices in the distribution
network.
Voltage Regulator
Voltage
Regulator
capacitor
banks
MCM
wireless
comm.
MIC
capacitor
banks
wireless
comm.
CERTS
Controller
Fig. 1.
wireless
comm.
CERTS
Controller
microgrid
CERTS
Controller
MIC
CERTS
Controller
CERTS
Controller
microgrid
microgrid
CERTS
Controller
MIC
CERTS
Controller
CERTS
Controller
CERTS
Controller
Complete system model of hierarchical system
To realize this objective, a hierarchical control architecture is proposed as illustrated in figure 1. At the basic
level of the multi-layer structure, we have microsource controllers tied to DERs within microgrids. These controllers
ensure the multi-unit stability of the network for a selected set point provided by their upper level microgrid controller
[5]. The MIC of a microgrid at the PCC designates a set point to each microsource controller satisfying PCC real
power and voltage conditions. For the microgrid consortium, the microgrid consortium manager (MCM) functions
as a supervisor. It determines status of the entire consortium, based on information from neighboring MICs. The
MCM does not directly designate set points to the coupled microgrids, but lets each microgrid determine its own
state. Each MIC maximizes its real power export through the PCC only using local information from its neighbors.
Compared to the centralized control structure, the decentralized approach places less burden on the communication
network, and it is also robust to system disturbances.
With the help of this hierarchical control architecture, the coupled microgrids provide not only real power service
but also voltage control service to the distribution network. Voltage control is a kind of “ancillary services” that must
be provided simultaneously with the transaction of power services [3]. Ancillary services are crucial for maintaining
reliable operation of the power grid. Besides voltage control service, one can also provide load following, operating
reserves, and energy imbalance with the same control system [4]. Then coupled microgrids would become a more
active player in both power service and ancillary service markets. This would encourage the integration of DERs
into the power grid.
Our work is only the first step of a comprehensive microgrid consortium controller. The next step is to let
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
3
the coupled microgrids achieve a common real or reactive power objective, with the same decentralized control
architecture. Because of the nonlinear and weak nature of the LV and MV distribution networks, this scheduling
problem is yet solved. Therefore, the microgrid consortium is capable of providing more ancillary services than
merely voltage control. Since both control layers are integrated in the MIC, this control architecture can be realized
R
with actual microgrid controllers from GE.
II. S YSTEM M ODEL
This section provides a detailed description of the distribution network model, as well as specifications and
assumptions used in our algorithm development, simulation, and verification processes. Properties of this network
are also identified, including its weak network connections and variable structure.
A. Distribution Network Model
Considering the hierarchical control architecture given in Section I, there are two layers of network model
corresponding to the MV distribution network and the LV microgrid respectively.
1) Model of MV Distribution Network: The distribution network model is shown in figure 2. There are N buses
in this network. Bus 1 is connected to the primary substation, hence represents the main grid. Microgrids or pure
loads connected to the remaining N − 1 buses are modeled as a microsource and a load tied to each bus. This
modification is appropriate, because microgrids keep their real and reactive power levels during operation.
Network Block
Bus 1
Bus 2
P1 Q1
Primary
Substation
P2 Q2
E 2 δ2
MicroSource Load
Fig. 2.
Bus 3
P3 Q3
E 3 δ3
. . . Bus N
. . .
MicroSource Load
PN QN
E N δN
MicroSource Load
MV distribution network model
Voltage magnitude and phase angle of bus i are Ei and δi . Specifically, bus 1 has E1 = 1.0 pu and δ1 = 0◦ .
Real and reactive power injected through bus i into the network are Pi and Qi . Signs of Pi and Qi represent the
direction of power flows, where a positive value means power injection into the network.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
4
2) Model of LV Microgrid: The microgrid controlled by the MICs is shown in figure 3. There are M buses in
this microgrid. Bus 1 is connected to the PCC, and represents the input/output port of the microgrid. A microsource
and a load are connected to each of the remaining (M − 1) buses.
P1 Q1
Bus 1
E 1 δ1
P2 Q2
Bus 2
E2 δ 2
Network
Block
E 3 δ3
E 4 δ4
P3 Q3
P4 Q4
Bus 3
Fig. 3.
Bus 4
LV microgrid model
Voltage magnitude and phase angle of bus i are Ei and δi . Real and reactive power injected through bus i into
the network are Pi and Qi . Signs of Pi and Qi represent the direction of power flows, where a positive value means
power injection into the microgrid. Specifically, bus 1 has E1 = Ei∗ and P1 = −Pi∗ , i.e. the state determined for
the PCC of the microgrid.
3) Expression of Real Power Loss: To express the real power loss Ploss over the connection lines, we define
the “network block” in figure 2 and figure 3. The “network block” only incorporates distribution lines connecting
all the buses. Because the block has no generation capability, it is a passive system and only capable of consuming
power. This definition helps when we consider the total line loss in our optimization problem. The real power loss
P
is the sum of real power injection into the network over all buses Ploss =
Pi .
B. Weak Network Property
Specifications of distribution lines is another very important aspect of our modeling. In contrast to HV transmission
lines, LV and MV distribution lines have a nontrivial ratio of
R
X,
where R is line resistance and X is line reactance.
This is usually called a “weak network”, indicating its vulnerability to disturbances.
More importantly, the real and reactive power are coupled together. The expressions of real and reactive power
export at bus i are as follows:
Pi
=
Ei
X
Ej (GBU S,ij cos(δi − δj ) + BBU S,ij sin(δi − δj ))
Qi
=
Ei
X
Ej (GBU S,ij sin(δi − δj ) − BBU S,ij cos(δi − δj ))
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
5
It is unable to linearize the equations above without making assumptions about the connections, i.e. GBU S,ij and
BBU S,ij in the expressions above. As a result, we consider weak systems with coupled real and reactive power
control in our analysis and algorithm development.
C. Structural Variation
In the electric power grid, structural variations are caused by switching operations. Because of the control and
protection operations conducted in distribution networks, the system structure is frequently changing. Such changes
have a direct impact on our hierarchical control system, since all set points are determined based on network
information.
Before responding to the structural changes, we have to detect them. Switchings happening within microgrids
are usually tractable by MICs, while switchings in MV distribution network are detectable by the MICs connected
by those corresponding links.
With the hierarchical control structure, responses to those structural changes can be confined to the corresponding
control layer. For instance, when the connection topology within a microgrid changes, it can be detected by its
MIC and used to update optimization problems, and then set point modifications can be designated. If the set point
before structural changes is still achievable by the microgrid, then disturbances will have no influence on system
operation beyond this microgrid.
III. A PPROACH
The hierarchical control system structure consists of low-level microsource controllers, mid-level MICs, and a
high-level MCM, as shown in figure 1. The control algorithm, however, is realized in the MICs. The objective is to
maximize the real power export from coupled microgrids with the proposed hierarchical control architecture. Three
aspects of problems are identified and solved in MICs, including decentralized control among coupled microgrids,
set point determination of microsources within microgrids, and microgrid capacity estimation.
A. Decentralized Control among Coupled Microgrids
The objective of the MIC optimization problem is to maximize real power export from this microgrid, and
minimize the real power loss along the distribution lines connected to it. The optimization subjects to constraints
on microgrid generation capacities, MV line power flow limits, regulatory constraints on voltage and frequency.
Solutions to this problem are the real power and voltage levels at the PCC. This set point can then be implemented
as constraints to solve another optimization problem to determine microsource set points. For bus i, assuming there
are M buses and L lines connecting them on the MV distribution network. This optimization problem is expressed
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
6
as:
M inimize
Pmg,i + Ploss,i
= Emg,i
M
X
Emg,j (GBU S,1j cos(δmg,1 − δmg,j ) + BBU S,1j sin(δmg,1 − δmg,j )) + Ploss,i
j=1
w.r.t.
subject
Emg,i
to
Pmg,i
Generation
Capability
Constraints
(i = 2, 3, · · · , M )
Pmg,i ≤ Pmg,i ≤ Pmg,i
Qmg,i ≤ Qmg,i ≤ Qmg,i
P ower
F low
Constraints
(i = 1, 2, · · · , L)
Pln,i ≤ Pln,i ≤ Pln,i
Qln,i ≤ Qln,i ≤ Qln,i
V oltage
Regulation
Rule
(i = 2, 3, · · · , M )
Emg,i ≤ Emg,i ≤ Emg,i
Pmg,i
=
Qmg,i
=
P ower Balance Relationship at P CC
X
Emg,i
Emg,j (GBU S,ij cos(δmg,i − δmg,j ) + BBU S,ij sin(δmg,i − δmg,j ))
X
Emg,i
Emg,j (GBU S,ij sin(δmg,i − δmg,j ) − BBU S,ij cos(δmg,i − δmg,j ))
In the optimization problem above, we need to compute Ploss,i and the power flowing into bus i based on set
points of its neighboring buses. The power flows from bus i to bus j and in the reverse direction are:
Pij
2
= −Emg,i
GBU S,ij + Emg,i Emg,j (GBU S,ij cos(δmg,i − δmg,j ) + BBU S,ij sin(δmg,i − δmg,j ))
Pji
2
= −Emg,j
GBU S,ji + Emg,j Emg,i (GBU S,ji cos(δmg,j − δmg,i ) + BBU S,ji sin(δmg,j − δmg,i ))
Based on the two power flow expressions, the real power loss along the line between bus i and bus j is Pij,loss =
2
2
Ploss,ij = | − GBU S,ij (Emg,i
− Emg,j
) + 2BBU S,ij sin(δmg,i − δmg,j )Emg,i Emg,j |. For the bus closest to the
primary substation, the line loss between it and the slack bus must be doubled. So that the sum of all the objective
functions within the distribution feeder would be the same as the one used in a centralized optimization problem.
By solving this optimization problem, the real power export to the distribution feeder is maximized, subject to
constraints on microgrid generation capacities, MV line power flow limits, regulatory constraint on voltage and
frequency. Since the microgrids are in close proximity, we can operate a communication network to exchange state
information between neighboring microgrids. The information required to solve this optimization problem is only
the voltage magnitude and phase-shift of directly connected microgrids.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
7
B. Microsource Set Points Determination
The MIC maintains the set point determined above, and determines set points of the microsources within its
microgrid. The set points are determined by solving an optimization problem, which aims to minimize the microgrid
running cost.
The objective of the MIC optimization problem is to minimize the generation costs while maintaining the given
set point at the PCC. The optimization is done subject to constraints on microsource generation capacities, LV
line power flow limits, regulatory constraints on voltage and frequency, and states at the PCC. Solutions to this
problem are the real power and voltage set points, as introduced in Section II. These set points are directly fed to
the microsource controllers as inputs. Assuming there are M buses and L lines connecting them, this optimization
problem is expressed as:
M inimize
C
M
X
!
Pms,i
i=2
w.r.t.
subject
Ei
to
(i = 2, 3, · · · , M )
Pms,i
Generation
Capability
Constraints
(i = 2, 3, · · · , M )
Pms,i ≤ Pms,i ≤ Pms,i
Qms,i ≤ Qms,i ≤ Qms,i
P ower
F low
Constraints
(i = 1, 2, · · · , L)
Pln,i ≤ Pln,i ≤ Pln,i
Qln,i ≤ Qln,i ≤ Qln,i
V oltage
Regulation
Rule
(i = 2, 3, · · · , M )
Ei ≤ Ei ≤ Ei
P ower Balance Relationship (i = 1, 2, · · · , M )
X
Ei
Ej (GBU S,ij cos(δi − δj ) + BBU S,ij sin(δi − δj ))
Pi
=
Pi
=
Qi
=
Pms,i − Pload,i
X
Ei
Ej (GBU S,ij sin(δi − δj ) − BBU S,ij cos(δi − δj ))
Qi
=
Qms,i − Qload,i
where C(·) is the cost function of running the microgrid, primarily the cost of power generation.
Since the microgrid concept organizes microsources in close proximity, we can operate a communication network
to transmit set points and to update topology and parameter changes. This optimization problem is solved numerically
in a centralized fashion, and set points are transmitted locally through a wireless communication network. The set
point information designated to each microsource includes real power and required voltage values.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
8
C. Microgrid Capacity Estimation
In the hierarchical control architecture, the microgrid capacities are used as constraints in the optimization
problems above, hence the real and reactive power capacity range must be identified for each microgrid. This
capacity is obtained by solving a series of optimization problems, and then selecting a subset of the applicable
region to ensure achievability.
Real and reactive power capacity must be determined over the voltage between 0.95pu and 1.05pu. First, the
reactive power range is determined, based on the optimization problem of maximizing and minimizing reactive power
export. At the extreme value of reactive power export, the maximum and minimum real power export capacities
are obtained from another set of optimization problems with such reactive power constraints.
To select the microgrid capacity achievable with any PCC voltage between 0.95pu and 1.05pu, we need to
assume that the region determined by the optimization problems above is convex. To simplify this procedure,
the PCC voltage is set to be 0.95pu, 1.0pu and 1.05pu. The real power range is determined by finding the
average maximum and minimum real power achievable with those three voltage levels. Although the power range
is conservative, it guarantees that every set point inside the region is achievable by the microgrid.
D. System Operation Procedure
With the system architecture shown in figure 1, a typical system operation procedure is described.
The MIC solves an optimization problem to maximize the real power injected by its microgrid through the PCC,
while considering the microgrid’s capacity, MV distribution line power flow constraints, and voltage regulation
along the line. The solution to this optimization problem is the state at the PCC. This state serves as a constraint for
another optimization problem also solved by the MICs, with constraints concerning the microsource’s capacity, line
flow limits within the microgrid, and regulatory constraints on voltage and frequency. This optimization problem
seeks to minimize the operational cost of the microgrid. Its output is a collection of voltage and real power set
points of microsources within the microgrid. These set points are transmitted to microsource controllers instantly.
The basic level microsource controllers use the real power and voltage set points as inputs, and actively adjust
network frequency and reactive power to maintain voltage and frequency stability during operation.
When component parameters or network topologies change, the microsource controllers should inform the MICs
about the changes that have happened. After updating its system model, the MIC recalculates the capacity of the
microgrid. The new capacity information is then sent to the MIC and used to recalculate the optimal solution for
the entire system.
IV. S IMULATION R ESULTS
This section provides the simulation results of the hierarchical control architecture in a distribution network with
coupled microgrids. In addition to steady state operations, system behaviors under structural disturbances are also
demonstrated. It is shown that the multi-layer control system is able to robustly respond to switching operations
happening in distribution networks.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
9
A. Simulation Model
The complete system architecture shown in figure 1 is integrated into the distribution network. In a three-bus
distribution network, each of the two microgrid buses accomodates a ∆-connected microgrid, as shown in figure 4.
Lines’ parameters are shown in table I.
ln 1
480 V
Primary
Substation
Load:
Microsource:
B1
ln 2
B3
B2
ln 7
ln 3
ln 4
30kW
ms 1
10kvar
0 ~ 60 kW
-20 ~ 20 kvar
ln 6
ms 2
Fig. 4.
ln 5 ln 8
ln 9
ms 4
ln 10
ms 3
ms 5
ms 6
Microgrid Simulation Model Schematic Model
TABLE I
L INE PARAMETERS OF D ISTRIBUTION N ETWORK
Length (mile)
R (Ω)
X (Ω)
480
1
0.786
0.156
480
0.5
0.393
0.078
AWG 1
480
0.2
0.1572
0.0312
AWG 4
480
0.2
0.316
0.035
ln 5
AWG 4
480
0.2
0.316
0.035
ln 6
AWG 4
480
0.2
0.316
0.035
ln 7
AWG 1
480
0.2
0.1572
0.0312
ln 8
AWG 4
480
0.2
0.316
0.035
ln 9
AWG 4
480
0.2
0.316
0.035
ln 10
AWG 4
480
0.2
0.316
0.035
Line
Type
Voltage (V)
ln 1
AWG 1
ln 2
AWG 1
ln 3
ln 4
Structural changes are also included in this simulation model. Each connection line is controlled by a switch,
hence there are various combinations of system topologies available. In this section, we will show the hierarchical
controller’s response to both structural changes within microgrids and on the lines connecting these microgrids.
The system not only maintains stable operation, but also converges to an optimal state.
The corresponding complete simulation model is built with SimPowerSystemsTM toolbox, as shown in figure 5.
At each bus within a microgrid, there is a microsource with real power capacity between 0kW and 60kW , and
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
10
reactive power capacity between −20kvar and 20kvar. In addition, a load is consuming 30kW real power and
10kvar reactive power.
S1_scope
Vabc
Iabc
S2_scope
cable_S1_to_S2
C
S3_scope
[S_ln2]
a
a
b
b
c
c
Vabc
A
com
A
A*
A
B
B*
B
C*
C
C
PQ
a
C
c
Discrete,
A
Vabc
b
b
C
PCC 1
c
A
Vabc
Iabc
PQ
a
The three-bus system is used to demonstrate how we can use solution
RMS (V-I)
B
cable_S2_to_S3
of the optimization problem as set point of the UWM microgrid controller.
Iabc
RMS (V-I)
B
Ts = 5e-005 s.
mg2_scope
C*
C
C
A
com
A
A*
b
B*
B
c
C*
C
C
a
certs_6_internal_scope
c
A
B
RMS (V-I)
Vabc
Iabc
PQ
b
c
B*
A*
C*
cable_certs5_to_certs6
b
[V_set3]
P_req
[P_set3]
certs_6_scope
B
E_req
m
C
a
c
A
m
C
A
B
A
C
B
certs_3_scope
a
b
c
B
C
A
A
B
C
[V_set6]
P_req
[P_set6]
C
A
-C-
ms 5
load_certs_6
B
b
c
a
A
com
-C-
B
-C-
C
C
C
A
B
b
A
B
C
C
B
load_certs_3
E_req
certs_6
ms 2
c
A
B
certs_3
a
a
B
b
c
B
A
C
C
B
Iabc
PQ
B
C*
C*
RMS (V-I)
B*
c
cable_certs2_to_certs3
com
[Smg2_1]
cable_certs4_to_certs6
com
A*
b
certs_3_internal_scope
a
Vabc
A
C
[Smg2_2]
A
a
a
B
b
A
cable_S3_to_certs4
C
A
com
com
B*
microgrid 2
[Smg1_1]
cable_certs1_to_certs3
[Smg1_2]
A*
A
microgrid 1
a
A*
cable_S2_to_certs1
B
A*
B*
mg1_scope
powergui
A
PCC 2
C
C
PQ
c
C*
C
Iabc
RMS (V-I)
B
C*
B
C
A
B*
c
A*
B
b
c
A
B
c
Vabc
A
com
a
b
b
a
b
PQ
a
RMS (V-I)
C
B*
C
Substation
[S_ln1]
RMS (V-I)
B
PQ
B
B
Iabc
A
A
[Smg2_4]
-C-
[Smg1_4]
[Smg1_3]
[Smg2_3]
A
com
B*
B
A
C*
C
B
[cap_mic2]
[trig_ln1]
[trig_ln2]
S_ln1
S_ln2
[Q_mg1]
[V_mag2]
V_pha2
[V_pha2]
recal_mg1
P_opt1
P_mg1
P_opt2
Q_mg1
P_opt3
[P_set3]
[V_mag2]
V_mag2
V_req_mg1
[V_set1]
change_mic3
[recal_mg2]
cap_mic3
P_opt_mic3
[P_mg2]
trig_ln2
Q_opt_mic3
[Q_mg2]
S_ln1
V_mag3
[V_mag3]
S_ln2
V_pha3
[V_pha3]
[switch_mg2]
[cap_mic3]
cap_est_mic3_snn
[V_pha2]
[switch_mg1]
[P_set1]
[P_set2]
V_pha2
V_req_mg2
[V_set2]
switch_mg1
V_req_mg3
[V_set3]
recal_mg2
P_opt1
[P_mg2]
P_mg2
P_opt2
[Q_mg2]
Q_mg2
P_opt3
[P_set6]
[V_mag3]
V_mag3
V_req_mg1
[V_set4]
[V_pha3]
V_pha3
V_req_mg2
[V_set5]
switch_mg2
V_req_mg3
[V_set6]
[recal_mg2]
[switch_mg2]
MIC1
Fig. 5.
Complete simulation model in Matlab with simpower toolbox
C
a
c
[P_set4]
-C-
A
B
0.6
A
B
C
V_mcm
[trig_mg2]
[P_mg1]
[V_set4]
P_req
ms 4
[cap_mic2]
[Q_mg1]
E_req
certs_4
cap_mic1
[recal_mg1]
b
c
C
B
A
[S_ln1]
cap_mic
PQ
b
-C-
ms 6
[trig_ln2]
V_mag2
B
A
Vabc
b
c
m
-Cload_certs_4
A
C
A
B
C
A
A
B
[P_set5]
load_certs_5
[cap_mic3]
cap_est_mic2_snn
[V_set5]
P_req
certs_5
[V_mcm]
[P_mg1]
Q_opt_mic2
[trig_mg1]
C
B
c
[S_ln2]
[switch_mg1]
E_req
a
b
[recal_mg1]
P_opt_mic2
trig_ln2
[S_ln2]
a
ms 1
trig_ln1
certs_4_internal_scope
certs_4_scope
m
-C-
cap_mic2
[S_ln1]
a
A
[P_set1]
C
B
C
change_mic2
[V_set1]
P_req
0.6
A
C
B
A
A
B
A
B
load_certs_1
V_mcm
E_req
certs_1
ms 3
[V_mcm]
Iabc
Vabc
b
-C-
C
load_certs_2
PQ
certs_5_scope
m
c
-C-
C
C
[P_set2]
B
C*
MIC2
B
[V_set2]
P_req
certs_2
B
c
B*
C
E_req
a
b
c
certs_5_internal_scope
C
a
RMS (V-I)
C
b
c
a
Vabc
Iabc
PQ
RMS (V-I)
B
A
C
c
a
b
A
B
RMS (V-I)
Vabc
Iabc
PQ
certs_1_internal_scope
certs_1_scope
m
b
cable_certs4_to_certs5
certs_2_internal_scope
certs_2_scope
A
A*
a
C
RMS (V-I)
A*
cable_certs1_to_certs2
Iabc
c
C
B
b
B
A
a
A
C
com
[P_set4]
[P_set5]
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
11
B. Steady State Operation
Our hierarchical control structure is capable of stabilizing the distribution network. The states of the system
converges to the steady state in less than one second, as shown in figure 6. The decentralized optimization problem
described in section III converges after 48 iterations, equivalent to 2.5ms. Since this is even less than one period
of AC electricity, performance of the decentralized system is satisfactory.
Microsource 1 Voltage
1000
0
−1000
0
0.5
1
1.5
Microsource 1 Current
2
2.5
0.5
1
1.5
Microsource 1 Real/Reactive Power
2
2.5
50
0
−50
0
4
x 10
0
P
Q
−2
−4
0
0.5
500
0
Fig. 6.
0
0.5
1
1.5
Voltage & Current (rms)
1
1.5
2
2.5
2
2.5
Microsource Steady State Scope
Figure 6 demonstrate the state of microsource “ms1”. At the beginning, voltage, current, real and reactive power
all show oscillations. After one second, the system states converge to the set point designated by MICs. The system
steady state is shown in figure 7. In addition to support all the local loads, the coupled microgrids export 33.22kW
real power to and require 75.38kvar reactive power from the main grid. With our hierarchical control system, the
coupled microgrids not only support local loads, but also provide real power services to the distribution network.
The total amount of real power export to the distribution network is less than the maximum real power that can
be extracted from the coupled microgrids. This reduction, however, is the result of our multi-layer architecture. A
microgrid is able to export more real power, but the capacity range in higher level algorithm may not be guaranteed.
With this sacrifice of microgrid capacity, system robustness can be improved. System operation under disturbances
are shown in the rest of this section.
C. Operations under Disturbances
Distribution networks are vulnerable to frequent switchings, and system stability may be jeopardized by these
typical operations. System response under network structural changes are considered in two aspects. The first case
is for structural changes within microgrids, and the second also includes changes to distribution lines connecting
the microgrids.
1) Changes within Microgrids: Changes are applied to the connection lines within microgrids. Line ln5 is
disconnected at t = 1s, and line ln8 is disconnected at t = 2s. Responses of the two microgrids are shown in
figure 8.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
P = 33.22 kW
Q = -75.38 kvar
Primary
Substation
492 V
1.025 pu
P = 30.63 kW
Q = 14.24 kvar
P = 30.63 kW
Q = 14.24 kvar
ms 1
60 kW
1.0414 pu
ms 2
30.63 kW
1.0381 pu
490.99 V
1.0229 pu
490.99 V
1.0229 pu
P = 28.53 kW
Q = -18.83 kvar
P = 28.53 kW
Q = -18.83 kvar
ms 4
ms 3
30.63 kW
1.0381 pu
60 kW
1.0157 pu
ms 6
ms 5
28.53 kW
1.0135 pu
28.53 kW
1.0135 pu
Distribution Network Steady State Simulation Result
Microgrid 1 Voltage
1000
0
2
4
6
50
8
10
12
Microgrid 1 Current
14
16
18
20
−1000
0
2
4
6
14
16
18
20
6
8
10
12
14
Microgrid 2 Real/Reactive Power
16
18
20
100
8
10
12
Microgrid 2 Current
0
0
−50
0
2
4
x 10
−1
4
6
8
10
12
14
Microgrid 1 Real/Reactive Power
16
18
−3
0
2
4
6
500
0
20
P
Q
−2
0
Microgrid 2 Voltage
1000
0
−1000
0
Fig. 8.
492.34 V
1.0257 pu
P = 60 kW
Q = - 20 kvar
494.88 V
1.031 pu
494.88 V
1.031 pu
Fig. 7.
P = 20.14 kW
Q = -89.55 kvar
495.07 V
1.0314 pu
P = 60 kW
Q = 20 kvar
Preq
Ereq
492 V
1.025 pu
P = 25.36 kW
Q = 16.61 kvar
480 V
1.0 pu
12
2
4
6
8
10
12
14
Voltage & Current (rms)
8
10
12
14
16
18
−100
0
2
5
x 10
1
4
P
Q
0
20
−1
0
2
4
6
2
4
6
500
16
18
20
0
0
8
10
12
14
Voltage & Current (rms)
8
10
12
14
16
18
20
16
18
20
Microgrid 1 and 2 Response to Changes within Microgrids
It is demonstrated that the real and reactive power injected by each microgrid change little after the switching
operations. Both the microgrids are stabilized after those changes within one second. Furthermore, the voltage
variations of participating microsources are all within 5% throughout the simulation process, as shown in figure 9.
As a result, the hierarchical control system is robust to structural changes within microgrids. In this simulation,
since the microgrid set point is still achievable after the disturbances, the MICs only reconfigure the interior of the
microgrid to maintain PCC state.
2) Changes Outside Microgrids: In addition to the changes discussed in the previous simulation, changes are
also applied to the connection lines between microgrids. Line “ln5” is disconnected at t = 1s, and line ”ln8” is
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
comparing rms voltages
500
450
400
350
400
350
300
250
250
200
200
150
150
100
100
50
50
Fig. 9.
2
4
6
8
10
12
14
16
18
mg 1
mg 2
450
300
0
0
comparing rms voltages
500
ms 1
ms 2
ms 3
ms 4
ms 5
ms 6
13
20
0
0
2
4
6
8
10
12
14
16
18
20
Voltage Responses to Structural Changes within Microgrids
disconnected at t = 5s. The line between “mg1” and “mg2” is disconnected at t = 10s. Responses of the two
microgrids are shown in figure 10.
Microgrid 1 Voltage
1000
Microgrid 2 Voltage
1000
0
0
−1000
0
2
4
6
100
8
10
12
Microgrid 1 Current
14
16
18
20
−1000
0
2
4
6
14
16
18
20
6
8
10
12
14
Microgrid 2 Real/Reactive Power
16
18
20
200
8
10
12
Microgrid 2 Current
0
0
−100
0
2
5
x 10
1
4
6
8
10
12
14
Microgrid 1 Real/Reactive Power
16
18
P
Q
0
−1
0
20
2
4
6
1000
8
10
12
14
Voltage & Current (rms)
16
18
−200
0
2
5
x 10
1
4
P
Q
0
20
−1
0
2
4
6
2
4
6
500
8
10
12
14
Voltage & Current (rms)
16
18
20
16
18
20
500
0
Fig. 10.
0
2
4
6
8
10
12
14
16
18
20
0
0
8
10
12
14
Microgrid 1 and 2 Response to Changes Outside Microgrids
It is demonstrated that the real and reactive power injected by each microgrid change fundamentally. After “ln2”
is disconnected, the two microgrids needs about six seconds to converge to the new set points designated by the
hierarchical control system, as shown in figure 11. From figure 10, the power export by microgrid “mg2” is reduced
to zero. Although the microgrid is disconnected from the network, it is still operating to support its local loads.
At the beginning of the system operation, there are 48 iterations of computation required to converge to the steady
state. After structural changes to connection lines between microgrids, the decentralized controller only needs to
exchange state information with neighboring agents once. Since the hierarchical control system converges soon after
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
comparing rms voltages
comparing rms voltages
500
ms 1
ms 2
ms 3
ms 4
ms 5
ms 6
450
400
350
500
400
350
300
250
250
200
200
150
150
100
100
50
50
Fig. 11.
0
2
4
6
8
10
12
14
16
18
mg 1
mg 2
450
300
0
14
20
0
0
2
4
6
8
10
12
14
16
18
20
Voltage Responses to Structural Changes Outside Microgrids
system variations, the communication burden of our controller is reasonable. More importantly, simulation results
show that the hierarchical control system is robust to structural changes in the distribution network, both inside and
outside of the microgrids.
V. C ONCLUSION AND F UTURE W ORK
In this report, a multi-layer hierarchical control system is proposed, which consists of basic level microsource
controllers, mid-level MICs, and a high level MCM. Microsource controllers ensure the multi-unit stability of
the network for a selected set point provided by their upper level microgrid controller. The MIC of a microgrid
connected to the PCC designates a set point to each microsource controller, satisfying the real power and voltage
level demanded. The MCM determines the operation status of the entire consortium, based on information from
neighboring MICs. The MIC not only determines its own state at the PCC, but also designate set points for
microsources within the microgrid. Compared to the centralized control structure, the decentralized approach places
less burden on communication networks, and it is also robust to disturbances in the distribution network.
With the help of this hierarchical control architecture, the coupled microgrids provide voltage control services
in addition to the real power services to the distribution network. This report is actually only the first step of a
comprehensive microgrid consortium controller. The next step is to let the coupled microgrids achieve a common
real or reactive power objective. Then besides voltage control service, one can also provide load following, operating
reserves, and energy imbalance services with the same control system. Then coupled microgrids would become a
more active player in both power service and ancillary service markets. Since both control layers are integrated in
the MIC, this control architecture can be realized with actual microgrid controllers from General Electric.
R EFERENCES
[1] Lasseter, R.H.; , ”Smart Distribution: Coupled Microgrids,” Proceedings of the IEEE , vol.99, no.6, pp.1074-1082, June 2011 doi:
10.1109/JPROC.2011.2114630.
DECENTRALIZED CONTROL ARCHITECTURE FOR COUPLING MICROGRIDS INTO DISTRIBUTION NETWORKS
15
[2] Masters, C.L.; , ”Voltage rise: the big issue when connecting embedded generation to long 11 kV overhead lines,” Power Engineering
Journal , vol.16, no.1, pp.5-12, Feb. 2002 doi: 10.1049/pe:20020101.
[3] Hirst, E.; Kirby, B.;, ”Electric-Power Ancillary Services,” Oak Ridge National Laboratory, Feb. 1996 ORNL/CON-426.
[4] ”Promotes Wholesale Competition through Open Access and Non-Discriminatory Transmission Service by Public Utilities,” Federal Energy
Regulatory Commission, RM95-8-000.
[5] Lasseter, R.; Piagi, P.; , ”Control and Design of Microgrid Components,” University of Wisconsin-Madison, January 2006.