30th Power System Conference (PSC2015), 23-25 November 2015, Niroo Research Institute, Tehran, Iran
Photovoltaic Microgrids Control by the Cooperative
Control of Multi-Agent Systems
H.A. Shayanfar, S.Malek**
Centre of Excellence for Power Automation and Operation, Electrical Engineering Department
Iran University of Science and Technology
Tehran, Iran
hashayanfar@yahoo.com,maleksajjad@yahoo.com
Abstract—This paper presents a cooperative control which is
applied to the secondary control of multi-agent controlled
microgrid. Balancing power in a microgrid is essential that leads
to voltage and frequency stabilities. The voltage and frequency
regulations are limiting them within specified limits and
approaching them to their nominal values. Limiting and
approaching voltage and frequency to their nominal values is
done by primary and secondary controls, respectively. Microgrid
has both dispatchable and non-dispatchable sources.
Dispatchable sources have been controlled by the conventional P-
w and Q-E droop controls. A photovoltaic as a non-dispatchable
source generates active power according to weather conditions,
but the reactive power supply is done by using E-Q droop
method. The E-Q droop uses the idle capacity of the inverters in
the reactive power sharing. Distributed secondary control
increases stability because of good, accurate and reliable
controls.Frequency is constant in the whole of microgrid. Since
line impedances are different, load terminal voltage control is
needed. The load as another agent is considered, who can request
desired voltage at its terminal bus.
Keywords—Cooperative Control, Multi-Agent System,
Secondary Control, Consensus Algorithm, Islanded Microgrid,
Photovoltaic System, Frequency Control, Voltage Control.
I. INTRODUCTION
Distributed Generation (DG) integration to distribution
grids has been done by microgrids. A microgrid is a set of
DGs that is controlled as a unit and is created in the purpose of
DGs management. Smart grids consist of several microgrids
which are named microgrid clusters and communicate with a
Distribution Management System (DMS).
Communication links are the smart grid essentials. A
reliable and high bandwidth communication link usually is
needed, especially when control is centralized. In contrary, a
decentralized control due to less communicated data, small
distance transmitted data and high reliability, is preferred.
Microgrids at the view of the connection to a main grid are
categorizedin two groups, are connected to the main grid or
islanded. Microgrids in normal operation are connected to the
main grid and exchange power with it, but when a fault in the
upstream grid occurs or power quality is very poor, microgrid
disconnects from it and is operated as a islanded microgrid. In
978-1-5090-2705-7/15/$31.00 ©2015 IEEE 287
islanded operating mode due to lack of a dominant source for
the voltage and frequency regulations, DGs cooperate in
voltage and frequency regulations and, as a result, active and
reactive powerssharing.
A hierarchical control scheme is used in microgrids.
Hierarchical structure includes primary, secondary and tertiary
levels. The primary level uses local data and measurements.
The secondary and tertiary levels need to communicate with
DGs. Both the secondary and tertiary levels can be centralized
or decentralized. In islanded mode, only the primary and
secondary controls are used and are responsible for the voltage
and frequency limiting and approaching them to the nominal
values.
Microgrid control by droop control had been used in [1, 2]
that causes voltage and frequency deviations and decreases
microgrid stability. An angle droop method had been
presented in [3, 4] that emulates the synchronous
generatorsbehavior and uses the synchronous generator
controls, but it is unable to regulate DC link voltage in
Photovoltaic (PV) systems. References [5-8]had proposed
central secondary control which need reliable communication
and complex structure. Centralized control development is
difficult and has low reliability. Distributed secondary controls
had been proposed in [9-12]that contains only ideal DC
sources. In [13] PV is controlled only in grid-connected mode
and islanded operating mode is not considered.
In this paper, PV Microgrid control by the cooperative
control of multi-agent system is proposed. The cooperative
control uses consensus algorithm for both voltage and
frequency. Load voltage must be kept in nominal value, so the
load is considered as an agent who can request desirable
voltage.
II. PRIMARY CONTROL LEVEL OF MICROGRID
In this section, the primary level control for the
dispatchable and non-dispatchable sources is addressed. The
dispatchable sources task is to regulate the voltage and
frequency and as a result, active and reactive powerscontrol.
Because the non-dispatchable sources like PV depends to the
weather conditions, therefore cannot control the output active
power, but by using the E − Q droop, the idle capacity of
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

inverters can be used in the reactive power supply in Where ΔE is the maximum allowed deviation and Q max is
cooperation with other sources. defined as the maximum available reactive power and is
A. Dispatchable Sources Control calculated as
The power stage and the control scheme of the
dispatchable sources are shown inFig. 1. The control is 2
Q max = S rated − Ppv2
designed in the synchronous rotating frame and dq (7)
transformation is used for transform line three phase
sinusoidal signals to constant signals. The control scheme Where S rated and Ppv are the rated PV inverter capacity and
includesP and Q calculation, droop control, voltage and the generated active power, respectively. Block diagram of the
current controllers. E − Q droop is illustrated inFig. 3.
The control scheme includes the P and Q calculation
which calculated the output active and reactive powers as III. SECONDARY CONTROL LEVEL
To overcome drawbacks of droop control, which causes
P = v od i od +v oq i oq (1) voltage and frequency deviations, the secondary control, is
proposed in[5-7]to restore them to their nominal values as
Q = −v od i oq +v oq i od (2)
ω = δω + ω* − m P P (8)
Where, P and Qare the output active and reactive powers,
respectively, v od and v oq are the components of voltage in d E = δ E + E * − nQQ (9)

and q axis, respectively, i od and i oq are the components of

current in d and q axis, respectively.
The well-known droop control is expressed as
ω = ω* − m P P
E = E * − n QQ
Where, ω and E are the frequency and voltage respectively,
ω* and E * are the frequency and voltage set points
respectively, m P and nQ are the droop coefficients
respectively.
The conventional proportional-integral controller is used in
voltage and current controllers. Further detailed information
can be found in [14].
Where δω and δ E are the signals which are generated by
the secondary control.
The secondary control due to communication with DGs
can be centralized or decentralized. The centralized and
(3) decentralized secondary controls are shown in Fig. 4 and Fig.
5, respectively. In this paper, decentralized control as multi-
agent systems is applied. The consensus algorithm is used as
(4)
the control protocol in this level and is introduced in next
section.
A. Consensus Algorithm for Secondary Control
1) Vf InvertersControl
The voltage and frequency regulations are done by these
type inverters. δ E for voltage control is generated as [9]
e Ei = e Load + ¦a
j ∈N i
ij (v odj − v odi ) + g i (E* − vodi ) + nQi Qi (10)

B. PV Control
The power stage and control system for a photovoltaic u Ei = c E e Ei (11)
system are shown inFig. 2. It includes E-Qdroop control, DC
link voltage, reactive power, and current controllers. A Phase δ E = ³ u Ei dt (12)
Locked Loop (PLL) is used for phase angle estimation. This
control system is in dqframe too. The E-Q droop is proposed
to use the idle capacity of PV inverters in the reactive power Where e Load is the generated error from close load which wants
supply and is expressed as[8] nominal voltage at its bus, N i is the set of neighbors of node i,
aij is the element of adjacency matrix, g i ≥ 0 is the pinning
E* −E
Q= gain and is nonzero at least for one bus, c ∈ \ and Q is
nQ E i
(5) calculated as

ΔE Qi = −ωcQ i + ωc (v oqi i odi − v odi i oqi ) (13)

nQ =
Q max
(6)

288
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

Fig. 1. Power Stage and Control System of Dispatchable Units[8]

Fig. 2. Power Stage and Control System of PV Systems[8]

289
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

Q max

E* −E
§ ΔE ·
¨ ¸
© Q max ¹

−Q max

Fig. 3. E − Q Droop Block Diagram[8]

δω

δE

Fig. 4. Centralized Microgrid Secondary Control in Islanded Mode[8]

e Load ω j , m j Pj ,v odj ωi , m i Pi ,v odi ,Qi

δE
e Ei = e Load + ¦a
j ∈N i
ij (v odj −v odi ) + g i (E* − vodi ) + nQi Qi u Ei = c E e Ei δ E = ³ u Ei dt

δω
eωi = ¦a ij (ω j − ωi ) + g i (ω * − ωi ) u ωi = c ω e ωi
j ∈N i
δω = ³ (u ωi + u Pi ) dt
e Pi = ¦a
j ∈N i
ij (m j Pj − m i Pi ) u Pi = c p e pi

e Load v odm v odk ,Q k

δE
e Ek = e Load + ¦a
m ∈N k
km (vodm − v odk ) + g k (E* − vodk ) + nQk Q k u Ek = c E e Ek δ E = ³ u Ek dt

Fig. 5. Multi-Agent Based Microgrid Secondary Control in Islanded Mode

290
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

Where ωc is the cut-off frequency of the low-pass filter (here found in [8]. cω , c P and c E are 400 and c Load = 20 are
is considered ωc = 2 × π × 5 ). For frequency regulation δω is considered as thecooperative control parameters. The g
calculated as [9] parameter is 1 only for DG-1 and is zero for other DGs.The
adjacency matrix is as
e ωi = ¦a ij (ω j − ωi ) + g i (ω * − ωi ) (14)
j ∈N i ª0 0 0 0º
«1 0 0 0 »»
A =« (23)
e Pi = ¦a ij (m j Pj − m i Pi ) (15) «1 0 0 0»
j ∈N i « »
¬0 0 1 0¼
u ωi = c ω e ω i (16)
To represent voltage regulation in microgrid The Voltage
Deviation Index is used and calculated as [15]
u Pi = c p e pi (17)
N Bus
(E i − E n )2
δω = ³ (u ωi + u Pi ) dt (18) V DI = ¦ E n2
i =1
(24)
Where cω , c P ∈ \ . Where Eι is the voltage at the bus i , E n is the nominal
2) PQ InvertersControl voltage (i.e. 380V), and N Bus is the number of buses.
In this agent, only voltage is regulated. So consensus
protocol is expressed as At the start of the simulation, PV and load agents are off.
At, t=0.5s the E-Qdroop of PVs are applied and PVs are
attended to the reactive power supply. It is seen that capacity
e Ei = e Load + ¦a
j ∈N i
ij (v odj − v odi ) + g i (E* − vodi ) + nQi Qi (19) of other Vf inverters gets release.At,t=1s the secondary control
is performed, and the voltage and frequency approaches to
their nominal values, so microgrid stability is improved. At,
u Ei = c E e Ei (20) t=1.5s the load-1 agent becomes active and sends appropriate
commands to DG-1 to regulate B-1 voltage in nominal value.
At, t=2s the load-2 agent becomes on and regulates its
δ E = ³ u Ei dt (21) terminal voltage in the nominal value. The active and reactive
powers are shown inFig. 7 andFig. 8, respectively. The three-
phase terminal voltage of DGs and voltage in each busis
3) Loads VoltageControl
shown in Fig. 9 and Fig. 10respectively. The angular
Critical Bus (CB) voltage regulation is important, so this
frequency of DG-1 and DG-3 is shown in Fig. 11.
agent is designed to control voltage magnitude on its terminal.
This agent action is expressed as The VDI in each time intervals is given in Table I. It can
be seen that voltage regulations in specific buses has a bad
e Load = c Load (E * − E Load ) (22) effect on other buses voltage. In order, load bus voltages get
better, but VDI in the microgrid gets worse.
Where E Load is load terminal voltage and c Load ∈ \ .

IV. SIMULATION RESULTS

In order to simulate the introduced control structures, a
microgrid which is shown inFig. 6, with four sources, is
considered. The Microgrid is simulated in MATLAB software.
The Microgrid has two dispatchable sources and two PV
arrays. The DG-1 and DG-3 are controlled by the P − ω and
Q − E droop control; DG-2 and DG-4 are controlled as
constant active power respect to the weather conditions and
E − Q droop. The consensus algorithm is used in the
secondary multi-agent system.Voltage and frequency
controlled inverters communicate each other and neighbor
power controlled inverters are connected to close voltage and
frequency controlled inverters. Also, loads as an
Fig. 6. Simulated microgrid structure
agentcommunicate with the close agent. Microgrid data can be

291
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

1.05

1.04

1.03

1.02

1.01

V
0.99 B1
V
B2

0.98 V
B3
V
B4
0.97

0.96
0.5 1 1.5 2 2.5 3
Time (s)

Fig. 7. Generated Active Power of DGs Fig. 10. Terminal Three-Phase Voltage of Buses

Fig. 8. Generated Reactive Power of DGs Fig. 11. Angular Frequency of DGs

1.06 TABLE I. VDI IN MICROGRID

V o1
V Time
1.05
V o3
o2

Interval
V B1 V B2 V B3 V B4 V DI
V
o4 0s-0.5s 378 379.6 368.7 381.3 0.000924
1.04
0.5s -1s 381.9 384.5 373.3 387.2 0.000835
1s-1.5s 391.2 393.1 386.9 398.9 0.004861
1.03
1.5s-2s 380.6 383.7 376.9 390.3 0.000899
2s-3s 380.6 384 379.8 392.7 0.001231
1.02

1.01
V. CONCLUSION
1 In this paper, a distributed secondary control based on the
cooperative control of the multi-agent systems for PV
0.99
0.5 1 1.5 2 2.5 3
Microgrids is proposed. The results show that the voltage and
Time (s) frequency in the microgrid approach to their nominal
values.As opposed tothe frequency which is constant in the
Fig. 9. Terminal Three-Phase Voltage of DGs
whole of microgrid, voltage is not constant in the whole of the
microgrid due to different line impedances. Hence, the load is
considered as an agent who can request desired voltage
magnitude. Load measures its terminal voltage and sends

292
 Photovoltaic Microgrids Control by the Cooperative Control of Multi-Agent Systems
30th Power System Conference - 2015 Tehran, Iran

theerror to a near agent and this agent tries to eliminate the [8] H. A. Shayanfar and S. Malek, "Islanded Photovoltaic Microgrid
error by adjusting voltage references. It can be concluded that Control With Proposed E-Q Droop and Secondary Control,"
presented at the International Conference and Exhibition on Solar
the loads voltage regulation has a bad effect on other buses. Energy (ICESE), Tehran, 2015.
Using load as an agent regulates its terminal voltage, but VDI [9] A. Bidram, A. Davoudi, F. L. Lewis, and Z. Qu, "Secondary
on microgrid buses gets worse. control of microgrids based on distributed cooperative control of
multi-agent systems," Generation, Transmission & Distribution,
REFERENCES IET, vol. 7, pp. 822-831, 2013.
[10] A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero,
[1] M. B. Delghavi and A. Yazdani, "Islanded-mode control of
"Distributed cooperative secondary control of microgrids using
electronically coupled distributed-resource units under unbalanced
feedback linearization," Power Systems, IEEE Transactions on,
and nonlinear load conditions," Power Delivery, IEEE
vol. 28, pp. 3462-3470, 2013.
Transactions on, vol. 26, pp. 661-673, 2011.
[11] P.-T. Cheng, C.-A. Chen, T.-L. Lee, and S.-Y. Kuo, "A
[2] J. W. Simpson-Porco, F. Dörfler, and F. Bullo, "Synchronization
cooperative imbalance compensation method for distributed-
and power sharing for droop-controlled inverters in islanded
generation interface converters," Industry Applications, IEEE
microgrids," Automatica, vol. 49, pp. 2603-2611, 2013.
Transactions on, vol. 45, pp. 805-815, 2009.
[3] M. Ashabani and Y. A.-R. Mohamed, "Integrating VSCs to weak
[12] Q. Shafiee, J. C. Vasquez, and J. M. Guerrero, "Distributed
grids by nonlinear power damping controller with self-
secondary control for islanded MicroGrids-A networked control
synchronization capability," Power Systems, IEEE Transactions
systems approach," in IECON 2012-38th Annual Conference on
on, vol. 29, pp. 805-814, 2014.
IEEE Industrial Electronics Society, 2012, pp. 5637-5642.
[4] M. Ashabani and Y. A.-R. Mohamed, "Novel comprehensive
[13] S. Bae and A. Kwasinski, "Dynamic modeling and operation
control framework for incorporating VSCs to smart power grids
strategy for a microgrid with wind and photovoltaic resources,"
using bidirectional synchronous-VSC," Power Systems, IEEE
Smart Grid, IEEE Transactions on, vol. 3, pp. 1867-1876, 2012.
Transactions on, vol. 29, pp. 943-957, 2014.
[14] N. Pogaku, M. Prodanoviü, and T. C. Green, "Modeling, analysis
[5] A. Bidram and A. Davoudi, "Hierarchical structure of microgrids
and testing of autonomous operation of an inverter-based
control system," Smart Grid, IEEE Transactions on, vol. 3, pp.
microgrid," Power Electronics, IEEE Transactions on, vol. 22, pp.
1963-1976, 2012.
613-625, 2007.
[6] J. M. Guerrero, J. C. Vasquez, J. Matas, D. Vicuna, L. García, and
[15] F. G. Montoya, R. Baños, C. Gil, A. Espín, A. Alcayde, and J.
M. Castilla, "Hierarchical control of droop-controlled AC and DC
Gómez, "Minimization of voltage deviation and power losses in
microgrids—A general approach toward standardization,"
power networks using Pareto optimization methods," Engineering
Industrial Electronics, IEEE Transactions on, vol. 58, pp. 158-172,
Applications of Artificial Intelligence, vol. 23, pp. 695-703, 2010.
2011.
[7] J. C. Vasquez, J. M. Guerrero, J. Miret, M. Castilla, D. Vicuna, and
L. García, "Hierarchical control of intelligent microgrids,"
Industrial Electronics Magazine, IEEE, vol. 4, pp. 23-29, 2010.

293