496 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 31, NO.

2, JUNE 2016

Control and Operation of a DC Grid-Based Wind

Power Generation System in a Microgrid
K. T. Tan, Member, IEEE, B. Sivaneasan, Member, IEEE, X. Y. Peng, Student Member, IEEE,
and P. L. So, Senior Member, IEEE

Abstract—This paper presents the design of a dc grid-based wind In recent years, the research attention on dc grids has been
power generation system in a poultry farm. The proposed system resurging due to technological advancements in power electron-
allows flexible operation of multiple parallel-connected wind gen- ics and energy storage devices, and increase in the variety of dc
erators by eliminating the need for voltage and frequency synchro-
nization. A model predictive control algorithm that offers better loads and the penetration of dc distributed energy resources
transient response with respect to the changes in the operating (DERs) such as solar photovoltaics and fuel cells.
conditions is proposed for the control of the inverters. The design Many research works on dc microgrids have been conducted
concept is verified through various test scenarios to demonstrate to facilitate the integration of various DERs and energy storage
the operational capability of the proposed microgrid when it oper- systems. In [5], [6], a dc microgrid based wind farm architecture
ates connected to and islanded from the distribution grid, and the
results obtained are discussed. in which each wind energy conversion unit consisting of a ma-
trix converter, a high frequency transformer and a single-phase
Index Terms—Wind power generation, dc grid, energy manage- ac/dc converter is proposed. However, the proposed architecture
ment, model predictive control.
increases the system complexity as three stages of conversion
are required. In [7], a dc microgrid based wind farm architec-
I. INTRODUCTION ture in which the WTs are clustered into groups of four with
each group connected to a converter is proposed. However, with
OULTRY farming is the raising of domesticated birds such
P as chickens and ducks for the purpose of farming meat or
eggs for food. To ensure that the poultries remain productive,
the proposed architecture, the failure of one converter will re-
sult in all four WTs of the same group to be out of service.
The research works conducted in [8]–[10] are focused on the
the poultry farms in Singapore are required to be maintained at
development of different distributed control strategies to coordi-
a comfortable temperature. Cooling fans, with power ratings of
nate the operation of various DERs and energy storage systems
tens of kilowatts, are usually installed to regulate the tempera-
in dc microgrids. These research works aim to overcome the
ture in the farms [1]–[3]. Besides cooling the farms, the wind
challenge of achieving a decentralized control operation using
energy produced by the cooling fans can be harnessed using
only local variables. However, the DERs in dc microgrids are
wind turbines (WTs) to reduce the farms’ demand on the grid.
strongly coupled to each other and there must be a minimum
The Singapore government is actively promoting this new con-
level of coordination between the DERs and the controllers. In
cept of harvesting wind energy from electric ventilation fans in
[11], [12], a hybrid ac/dc grid architecture that consists of both ac
poultry farms which has been implemented in many countries
and dc networks connected together by a bidirectional converter
around the world [4]. The major difference between the situa-
is proposed. Hierarchical control algorithms are incorporated
tion in poultry farms and common wind farms is in the wind
to ensure smooth power transfer between the ac microgrid and
speed variability. The variability of wind speed in wind farms
the dc microgrid under various operating conditions. However,
directly depends on the environmental and weather conditions
failure of the bidirectional converter will result in the isolation
while the wind speed in poultry farms is generally stable as it is
of the dc microgrid from the ac microgrid.
generated by constant-speed ventilation fans. Thus, the genera-
An alternative solution using a dc grid based distribution
tion intermittency issues that affect the reliability of electricity
network where the ac outputs of the wind generators (WGs)
supply and power balance are not prevalent in poultry farm wind
in a poultry farm are rectified to a common voltage at the dc
energy systems.
grid is proposed in this paper. The most significant advantage of
the proposed system is that only the voltage at the dc grid has
Manuscript received February 7, 2015; revised April 27, 2015 and August to be controlled for parallel operation of several WGs without
8, 2015; accepted October 27, 2015. Date of publication December 22, 2015; the need to synchronize the voltage, frequency and phase, thus
date of current version May 18, 2016. This work was supported by the School allowing the WGs to be turned ON or OFF anytime without
of Electrical and Electronic Engineering, Nanyang Technological University,
Singapore, and the Electrical Engineering Division, Ngee Ann Polytechnic, causing any disruptions.
Singapore. Paper no. TEC-00126-2015. Many research works on designing the controllers for the
K. T. Tan is with the Electrical Engineering Division, Ngee Ann Polytechnic, control of inverters in a microgrid during grid-connected and
Singapore (e-mail: [email protected]).
B. Sivaneasan is with the School of Engineering, Nanyang Polytechnic, islanded operations is conducted in [13]–[15]. A commonly
Singapore (e-mail: [email protected]). adopted control scheme which is detailed in [13], [14] contains
X. Y. Peng and P. L. So are with the School of Electrical and Electronic an inner voltage and current loop and an external power loop to
Engineering, Nanyang Technological University, Singapore (e-mail: xypeng2@
e.ntu.edu.sg; [email protected]). regulate the output voltage and the power flow of the inverters.
Digital Object Identifier 10.1109/TEC.2015.2497709 In [15], a control scheme which uses separate controllers for
0885-8969 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
TAN et al.: CONTROL AND OPERATION OF A DC GRID-BASED WIND POWER GENERATION SYSTEM IN A MICROGRID 497

the inverters during grid-connected and islanded operations is II. SYSTEM DESCRIPTION AND MODELING
proposed. Although there are a lot of research works being con-
A. System Description
ducted on the development of primary control strategies for DG
units, there are many areas that require further improvement and The overall configuration of the proposed dc grid based wind
research attention. These areas include improving the robustness power generation system for the poultry farm is shown in Fig. 1.
of the controllers to topological and parametric uncertainties, The system can operate either connected to or islanded from
and improving the transient response of the controllers. the distribution grid and consists of four 10 kW permanent
To increase the controller’s robustness against variations in magnet synchronous generators (PMSGs) which are driven by
the operating conditions when the microgrid operates in the the variable speed WTs. The PMSG is considered in this paper
grid-connected or islanded mode of operation as well as its ca- because it does not require a dc excitation system that will
pability to handle constraints, a model-based model predictive increase the design complexity of the control hardware. The
control (MPC) design is proposed in this paper for controlling three-phase output of each PMSG is connected to a three-phase
the inverters. As the microgrid is required to operate stably in converter (i.e., converters A, B, C and D), which operates as a
different operating conditions, the deployment of MPC for the rectifier to regulate the dc output voltage of each PMSG to the
control of the inverters offers better transient response with re- desired level at the dc grid. The aggregated power at the dc grid
spect to the changes in the operating conditions and ensures is inverted by two inverters (i.e., inverters 1 and 2) with each
a more robust microgrid operation. There are some research rated at 40 kW. Instead of using individual inverter at the output
works on the implementation of MPC for the control of invert- of each WG, the use of two inverters between the dc grid and
ers. In [16], a finite control set MPC scheme which allows for the ac grid is proposed. This architecture minimizes the need to
the control of different converters without the need of additional synchronize the frequency, voltage and phase, reduces the need
modulation techniques or internal cascade control loops is pre- for multiple inverters at the generation side, and provides the
sented but the research work does not consider parallel operation flexibility for the plug and play connection of WGs to the dc
of power converters. In [17], an investigation on the usefulness grid. The availability of the dc grid will also enable the supply
of the MPC in the control of parallel-connected inverters is con- of power to dc loads more efficiently by reducing another ac/dc
ducted. The research work is, however, focused mainly on the conversion.
control of inverters for uninterruptible power supplies in stand- The coordination of the converters and inverters is achieved
alone operation. The MPC algorithm will operate the inverters through a centralized energy management system (EMS). The
close to their operating limits to achieve a more superior perfor- EMS controls and monitors the power dispatch by each WG
mance as compared to other control methods which are usually and the load power consumption in the microgrid through a
conservative in handling constraints [18], [19]. In this paper, centralized server. To prevent excessive circulating currents be-
the inverters are controlled to track periodic current and volt- tween the inverters, the inverter output voltages of inverters 1
age references and the control signals have a limited operating and 2 are regulated to the same voltage. Through the EMS, the
range. Under such operating condition, the MPC algorithm is output voltages of inverters 1 and 2 are continuously monitored
operating close to its operating limits where the constraints will to ensure that the inverters maintain the same output voltages.
be triggered repetitively. In conventional practices, the control The centralized EMS is also responsible for other aspects of
signals are clipped to stay within the constraints, thus the sys- power management such as load forecasting, unit commitment,
tem will operate at the sub-optimal point. This results in inferior economic dispatch and optimum power flow. Important infor-
performance and increases the steady-state loss. MPC, on the mation such as field measurements from smart meters, trans-
contrary, tends to make the closed-loop system operate near its former tap positions and circuit breaker status are all sent to
limits and hence produces far better performance. MPC has also the centralized server for processing through wireline/wireless
been receiving increased research attention for its applications communication. During normal operation, the two inverters will
in energy management of microgrids because it is a multi-input, share the maximum output from the PMSGs (i.e., each inverter
multi-output control method and allows for the implementation shares 20 kW). The maximum power generated by each WT
of control actions that predict future events such as variations in is estimated from the optimal wind power Pwt,opt as follows
power generation by intermittent DERs, energy prices and load [23]:
demands [20]–[22]. In these research works, the management
Pwt,opt = kopt (ωr,opt )3 (1)
of energy is formulated into different multi-objective optimiza-
tion problems and different MPC strategies are proposed to  3
1 R
solve these optimization problems. The scope of this paper is kopt = Cp,opt ρA (2)
2 λopt
however focused on the application of MPC for the control of
inverters. λopt v
ωr,opt = (3)
In what follows, a comprehensive solution for the operation of R
a dc grid based wind power generation system in a microgrid is where kopt is the optimized constant, ωr,opt is the WT speed
proposed for a poultry farm and the effectiveness of the proposed for optimum power generation, Cp,opt is the optimum power
system is verified by simulation studies under different operating coefficient of the turbine, ρ is the air density, A is the area swept
conditions. by the rotor blades, λopt is the optimum tip speed ratio, v is the
498 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 31, NO. 2, JUNE 2016

Fig. 1. Overall configuration of the proposed dc grid based wind power generation system in a microgrid.

wind speed and R is the radius of the blade. When one inverter B. System Operation
fails to operate or is under maintenance, the other inverter can When the microgrid is operating connected to the distribution
handle the maximum power output of 40 kW from the PMSGs. grid, the WTs in the microgrid are responsible for providing
Thus the proposed topology offers increased reliability and en-
local power support to the loads, thus reducing the burden of
sures continuous operation of the wind power generation system power delivered from the grid. The SB can be controlled to
when either inverter 1 or inverter 2 is disconnected from opera- achieve different demand side management functions such as
tion. An 80 Ah storage battery (SB), which is sized according to
peak shaving and valley filling depending on the time-of-use of
[24], is connected to the dc grid through a 40 kW bidirectional electricity and SOC of the SB [27]–[29].
dc/dc buck-boost converter to facilitate the charging and dis- During islanded operation where the CBs disconnect the mi-
charging operations when the microgrid operates connected to
crogrid from the distribution grid, the WTs and the SB are only
or islanded from the grid. The energy constraints of the SB in the available sources to supply the load demand. The SB can supply
proposed dc grid are determined based on the system-on-a-chip
for the deficit in real power to maintain the power balance of the
(SOC) limits given by
microgrid as follows:
SOCm in < SOC ≤ SOCm ax (4)
Pwt + Psb = Ploss + Pl (5)
Although the SOC of the SB cannot be directly measured, it
can be determined through the estimation methods as detailed where Pwt is the real power generated by the WTs, Psb is the
in [25], [26]. With the use of a dc grid, the impact of fluctuations real power supplied by SB which is subjected to the constraint
between power generation and demand can be reduced as the of the SB maximum power Psb,m ax that can be delivered during
SB can swiftly come online to regulate the voltage at the dc grid. discharging and is given by
During off-peak periods when the electricity demand is low, the
SB is charged up by the excess power generated by the WTs. Psb ≤ Psb,m ax (6)
Conversely, during peak periods when the electricity demand is
high, the SB will supplement the generation of the WTs to the Ploss is the system loss, and Pl is the real power that is
loads. supplied to the loads.
TAN et al.: CONTROL AND OPERATION OF A DC GRID-BASED WIND POWER GENERATION SYSTEM IN A MICROGRID 499

Fig. 2. Power circuit of a PMSG connected to an ac/dc voltage source con-

verter. Fig. 3. Single-phase representation of the three-phase dc/ac inverter.

C. AC/DC Converter Modeling

current, i is the current flowing through Lf , iC f is the current
Fig. 2 shows the power circuit consisting of a PMSG which flowing through Cf , and vDG is the inverter output voltage.
is connected to an ac/dc voltage source converter. The PMSG is During grid-connected operation, the inverters are connected
modeled as a balanced three-phase ac voltage source esa , esb , esc to the distribution grid and are operated in the current control
with series resistance Rs and inductance Ls [30], [31]. As shown mode (CCM) because the magnitude and the frequency of the
in [32], the state equations for the PMSG currents isa , isb , isc output voltage are tied to the grid voltage. Thus, the discrete
and the dc output voltage Vdc of the converter can be expressed state-space equations for the inverter model operating in the
as follows: CCM can be expressed with sampling time Ts as follows:
dis
Ls = − Rs is + es − KSVdc (7) xg (k + 1) = Ag xg (k) + Bg 1 vg (k) + Bg 2 ug (k) (12)
dt
dVdc yg (k) = Cg xg (k) + Dg vg (k) (13)
C = iTs S − Idc (8)
dt
where where the subscript g represents the inverter model during grid-
 T  T connected operation, k is the discretized present time step, and
is = isa isb isc , es = esa esb esc
⎡ ⎤ R  Vdc
2/3 −1/3 −1/3 Ag = 1 − Ts , Bg 1 = 0 − LT sf , Bg 2 = Ts
⎢ ⎥ Lf Lf
K= ⎢ ⎣ −1/3 2/3 −1/3 ⎥
⎦ 
Cg = 1, Dg = CT sf − CT sf
−1/3 −1/3 2/3
 T
S = Sa S b S c is the ac/dc converter switching functions xg (k) = i(k) is the state vector; vg (k) = [vDG (k + 1)
which are defined as vDG (k)]T is the exogenous input; ug (k) is the control signal
 with −1 ≤ ug (k) ≤ 1; and yg (k) = iDG (k) is the output. The
1, Sj is ON
Sj = for j = a, b, c (9) exogenous input vg (k) can be calculated using state estimation.
0, Sj is OFF In this paper, the grid is set as a large power system, which means
that the grid voltage is a stable three-phase sinusoidal voltage.
D. DC/AC Inverter Modeling Hence, when operating in the CCM, a three-phase sinusoidal
The two 40 kW three-phase dc/ac inverters which connect the signal can be used directly as the exogenous input.
dc grid to the point of common coupling (PCC) are identical, and During islanded operation, the inverters will be operated in
the single-phase representation of the three-phase dc/ac inverter the voltage control mode (VCM). The voltage of the PCC will
is shown in Fig. 3. To derive a state-space model for the inverter, be maintained by the inverters when the microgrid is islanded
Kirchhoff’s voltage and current laws are applied to loop i and from the grid. As compared to Ts , the rate of change of the
point x respectively, and the following equations are obtained: inverter output current is much slower. Therefore, the following
assumption is made when deriving the state-space equations for
di the inverter operating in the VCM [33]:
Lf + iR + vDG = uVdc (10)
dt
iDG = i − iC f (11) diDG
= 0. (14)
dt
where Vdc is the dc grid voltage, u is the control signal, R is the
inverter loss, Lf and Cf are the inductance and capacitance of Based on the above mentioned assumption, the discrete state-
the low-pass (LPF) filter respectively, iDG is the inverter output space equations of the inverter model operating in the VCM can
500 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 31, NO. 2, JUNE 2016

B. Control Design for the DC/AC Inverter

In order for the microgrid to operate in both grid-connected
and islanded modes of operation, a model-based controller us-
ing MPC is proposed for the control of the inverters. MPC is a
model-based controller and adopts a receding horizon approach
in which the optimization algorithm will compute a sequence
of control actions to minimize the selected objectives for the
whole control horizon, but only execute the first control action
for the inverter. At the next time step, the optimization process
is repeated based on new measurements over a shifted predic-
tion horizon. By doing so, MPC can make the output track
the reference at the next step, as well as plan and correct its
Fig. 4. Configuration of the proposed controller for the ac/dc converter. control signals along the control process. This will guarantee a
better transient response compared to conventional PID/PR con-
trollers [32], [33]. To derive the control algorithm for the invert-
be expressed as follows: ers, the state-space equations are transformed into augmented
state-space equations by defining the incremental variables in
xi (k + 1) = Ai xi (k) + Bi ui (k) (15) the following format:
yi (k) = Ci xi (k) (16)
Δξ(k) = ξ(k) − ξ(k − 1) (17)
where the subscript i represents the model of the inverter during
islanded operation and where ξ represents each variable in the inverter model, such as
vDG , iDG , i and u as shown in Fig. 3.
⎡ ⎤
R Ts By defining the incremental variables, the augmented state-
⎢ 1 − T s − 0 ⎥
Lf Lf   space model for the inverter model operating in the CCM during
⎢ ⎥
⎢T ⎥ Vdc grid-connected operation can be expressed as follows:
Ai = ⎢ s Ts ⎥ , Bi = Ts 0 0
⎢ 1 − ⎥ Lf
⎣ Lf Cf ⎦
Xg (k + 1) = Ag aug Xg (k)
0 0 1
  + Bg 1 aug Vg (k) + Bg 2 aug Ug (k) (18)
Ci = 0 1 0
Yg (k) = Cg aug Xg (k) (19)
 T
xi (k) = i(k) vDG (k) iDG (k) is the state vector; ui (k) is
the control signal with −1 ≤ ui (k) ≤ 1; and yi (k) = vDG (k) is where
the output. During islanded operation, the inverters are required ⎡ ⎤
R
to deliver all the available power from the PMSGs to the loads. 1− Ts 0
⎢ Lf ⎥
Therefore, only the inverter output voltage is controlled and ⎢ ⎥
Ag aug = ⎢ ⎥,
the output current is determined from the amount of available ⎣ R ⎦
power. 1− Ts 1
Lf
⎡ ⎤
Ts
III. CONTROL DESIGN 0 0 −
⎢ Lf ⎥
⎢ ⎥
A. Control Design for the AC/DC Converter Bg 1 aug = ⎢ ⎥
⎣ Cf Cf Ts ⎦
Fig. 4 shows the configuration of the proposed controller − −
Ts Ts Lf
for each ac/dc voltage source converter which is employed to ⎡ ⎤
maintain the dc output voltage Vdc of each converter and com- Vdc Vdc
T − Ts
⎢ Lf s Lf ⎥  
pensate for any variation in Vdc due to any power imbalance in ⎢ ⎥
the dc grid. The power imbalance will induce a voltage error Bg 2 aug = ⎢ ⎥ , Cg aug = 0 1
⎣ Vdc Vdc ⎦
∗ −
(Vdc − Vdc ) at the dc grid, which is then fed into a proportional- Ts Ts
Lf Lf
integral controller to generate a current reference i∗d for id to
track. To eliminate the presence of high frequency switching  T
ripples at the dc grid, Vdc is first passed through a first-order Xg (k) = Δi(k) iDG (k) is the state vector; Vg (k) =
 T
LPF. The current iq is controlled to be zero so that the PMSG ΔvDG (k + 2) ΔvDG (k + 1) ΔvDG (k) is the exogenous
only delivers real power. The current errors Δid and Δiq are input; Ug (k) = Δug (k) is the control signal; and Yg (k) =
then converted into the abc frame and fed into a proportional- iDG (k) is the output.
resonant (PR) controller to generate the required control signals Similarly, the augmented state-space model for the inverter
using pulse-width modulation. model operating in the VCM during islanded operation can be
TAN et al.: CONTROL AND OPERATION OF A DC GRID-BASED WIND POWER GENERATION SYSTEM IN A MICROGRID 501

expressed as follows: TABLE I

PARAMETERS OF THE PROPOSED SYSTEM
Xi (k + 1) = Ai aug Xi (k) + Bi aug Ui (k) (20)
Parameter Value
Yi (k) = Ci aug Xi (k) (21)
Distribution grid voltage v g = 230 V (phase)
where DC grid voltage V d c = 500 V
PMSG stator impedance R s = 0.2 Ω, L s = 2.4 mH
⎡ ⎤ Distribution line impedance R  = 7.5 mΩ, L  = 25.7 μH
R Ts
1− Ts − 0 0 Inverter LC filter L f = 1.2 mH, C f = 20 μF
⎢ Lf Lf ⎥ Converter capacitor C = 300 μF
⎢ ⎥
⎢ ⎥ Converter and inverter loss resistance R = 1 mΩ
⎢ Ts Ts ⎥ Load 1 rating P L 1 = 35 kW, Q L 1 = 8 kVAr
⎢ − 0⎥
⎢L 1 ⎥ Load 2 rating P L 2 = 25 kW, Q L 2 = 4 kVAr
⎢ f Lf ⎥
Ai aug = ⎢ ⎥,
⎢ ⎥
⎢0 0 1 0 ⎥
⎢ ⎥
⎢ ⎥ MATLAB/Simulink. The effectiveness of the proposed design
⎢ ⎥
⎣ Ts Ts ⎦ concept is evaluated under different operating conditions when
1 1 the microgrid is operating in the grid-connected or islanded
Lf Lf
⎡ ⎤ mode of operation. The system parameters are given in Table I.
Vdc The impedances of the distribution line are obtained from [34].
⎢ Lf Ts ⎥
⎢ ⎥ In practical implementations, the values of the converter and
⎢ ⎥
⎢ 0 ⎥ inverter loss resistance are not precisely known. Therefore, these
⎢ ⎥  
Bi aug = ⎢ ⎢
⎥ , Ci aug = 0 0 0 1
⎥
values have been coarsely estimated.
⎢ ⎥
⎢ 0 ⎥
⎢ ⎥ A. Test Case 1: Failure of One Inverter During
⎣ ⎦
Grid-Connected Operation
0
When the microgrid is operating in the grid-connected mode
 T
Xi (k) = Δi(k) ΔvDG (k) ΔiDG (k) vDG (k) is the state of operation, the proposed wind power generation system will
vector; Ui (k) = Δui (k) is the control signal; and Yi (k) = supply power to meet part of the load demand. Under normal
vDG (k) is the output. operating condition, the total power generated by the PMSGs
For the control of the two augmented models in the CCM and at the dc grid is converted by inverters 1 and 2 which will
the VCM, the following cost function is solved using quadratic share the total power supplied to the loads. When one of the
programming in the proposed MPC algorithm [33]: inverters fails to operate and needs to be disconnected from the
dc grid, the other inverter is required to handle all the power
J = (Rs − Yj )T (Rs − Yj ) + UjT QUj (22) generated by the PMSGs. In this test case, an analysis on the
subject to the constraint microgrid operation when one of the inverters is disconnected
from operation is conducted.
−1 ≤ uj (k) ≤ 1 (23) With each PMSG generating about 5.5 kW of real power, the
total power generated by the four PMSGs is about 22 kW which
where Rs is the set-point matrix, Q is the tuning matrix for the
is converted by inverters 1 and 2 into 20 kW and 8 kVAr of real
desired closed-loop performance, Yj is the output of either the
and reactive power respectively. Figs. 5 and 6 show the wave-
augmented model in the CCM or VCM (i.e., Yg or Yi ), Uj is
forms of the real and reactive power delivered by inverters 1 and
the control signal of either the augmented model in the CCM
2 for 0 ≤ t < 0.4 s respectively. For 0 ≤ t < 0.2 s, both invert-
or VCM (i.e., Ug or Ui ). The first part of the cost function
ers 1 and 2 are in operation and each inverter delivers about 10
is to compare the output of the augmented model Yj with the
kW of real power and 4 kVAr of reactive power to the loads.
reference Rs and to ensure that the output tracks the reference
The remaining real and reactive power that is demanded by the
with minimum error. The second part of the cost function is to
loads is supplied by the grid which is shown in Fig. 7. It can be
calculate the weighted factor of the control signal and to ensure
seen from Fig. 7 that the grid delivers 40 kW of real power and
that the control signal generated by the MPC algorithm is within
4 kVAr of reactive power to the loads for 0 ≤ t < 0.2 s. The
the constraints. The quadratic programming will ensure that the
total real and reactive power supplied to the loads is about 60
optimal solution for the control signal deviation Δu is achieved
kW and 12 kVAr as shown in the power waveforms of Fig. 8.
while minimizing the cost function J. After the control signal
The unsteady measurements observed in the power waveforms
u is generated by the MPC algorithm, it will be applied to the
for 0 ≤ t < 0.08 s are because the controller requires a pe-
dc/ac inverter as shown in Fig. 3.
riod of about four cycles to track the power references during
the initialization period. As compared to conventional control
IV. NUMERICAL SIMULATION ANALYSIS
strategies, it can be observed that the proposed MPC algorithm
The simulation model of the proposed dc grid based wind is able to quickly track and settle to the power reference. This
power generation system shown in Fig. 1 is implemented in is attributed to the optimization of the inverters through the
502 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 31, NO. 2, JUNE 2016

Fig. 5. Real (top) and reactive (bottom) power delivered by inverter 1. Fig. 8. Real (top) and reactive (bottom) power consumed by the loads.

Fig. 9. DC grid voltage.

in Fig. 5, the real and reactive power supplied by inverter 1 is

decreased to zero in about half a cycle after inverter 1 is discon-
nected. This undelivered power causes a sudden power surge in
the dc grid which corresponds to a voltage rise at t = 0.2 s as
Fig. 6. Real (top) and reactive (bottom) power delivered by inverter 2.
shown in Fig. 9. To ensure that the load demand is met, the grid
automatically increases its real and reactive power generation to
50 kW and 8 kVAr respectively at t = 0.2 s, as shown in Fig. 7.
At t = 0.26 s, the EMS of the microgrid increases the reference
real and reactive power supplied by inverter 2 to 20 kW and
8 kVAr respectively. A delay of three cycles is introduced to
cater for the response time of the EMS to the loss of inverter
1. As shown in Fig. 6, inverter 2 manages to increase its real
and reactive power supplied to the loads to 20 kW and 8 kVAr
for 0.26 ≤ t < 0.4 s. At the same time, the grid decreases its
real and reactive power back to 40 kW and 4 kVAr as shown
in Fig. 7 respectively. The power balance in the microgrid is
restored after three cycles from t = 0.26 s. It is observed from
Fig. 9 that the voltage at the dc grid corresponds to a voltage dip
at t = 0.26 s due to the increase in power drawn by inverter 2 and
then returns to its nominal value of 500 V for 0.26 ≤ t < 0.4 s.
Fig. 7. Real (top) and reactive (bottom) power delivered by the grid.
As observed in Fig. 8, at t = 0.26 s, the changes in power deliv-
ered by inverter 2 and the grid also cause a transient in the load
power.
model-based MPC control. Essentially, model-based control
schemes are able to take into account the system parameters
such that the overall performance can be optimized. B. Test Case 2: Connection of AC/DC Converter During
Grid-Connected Operation
At t = 0.2 s, inverter 1 fails to operate and is disconnected
from the microgrid, resulting in a loss of 10 kW of real power The most significant advantage of the proposed dc grid based
and 4 kVAr of reactive power supplied to the loads. As shown wind power generation system is that it facilitates the connection
TAN et al.: CONTROL AND OPERATION OF A DC GRID-BASED WIND POWER GENERATION SYSTEM IN A MICROGRID 503

Fig. 10. Real (top) and reactive (bottom) power delivered by inverter 1. Fig. 12. Real (top) and reactive (bottom) power delivered by the grid.

Fig. 13. DC grid voltage.

dip in the dc grid voltage at t = 0.26 s as observed in Fig. 13

which is then restored back to its nominal voltage of 500 V
for 0.26 ≤ t < 0.4 s. The grid also simultaneously decreases its
supply to 40 kW of real power for 0.26 ≤ t < 0.4 s while its
Fig. 11. Real (top) and reactive (bottom) power delivered by inverter 2.
reactive power remains constant at 4 kVAr as shown in Fig. 12.

C. Test Case 3: Islanded Operation

of any PMSGs to the microgrid without the need to synchronize
their voltage and frequency. This capability is demonstrated in When the microgrid operates islanded from the distribution
this case study. grid, the total generation from the PMSGs will be insuffi-
The microgrid operates connected to the grid and PMSG A cient to supply for all the load demand. Under this condition,
is disconnected from the dc grid for 0 ≤ t < 0.2 s as shown in the SB is required to dispatch the necessary power to ensure
Fig. 1. The real power generated from each of the remaining that the microgrid continues to operate stably. The third case
three PMSGs is maintained at 5.5 kW and their aggregated real study shows the microgrid operation when it islands from the
power of 16.5 kW at the dc grid is converted by inverters 1 and grid.
2 into 14 kW of real power and 8 kVAr of reactive power. As The microgrid is initially operating in the grid-connected
shown in Figs. 10 and 11, each inverter delivers real and reactive mode. The grid is supplying real power of 40 kW and reactive
power of 7 kW and 4 kVAr to the loads respectively. The rest of power of 4 kVAr to the loads for 0 ≤ t < 0.2 s as shown in
the real and reactive power demand of the loads is supplied by Fig. 14 while each inverter is delivering real power of 10 kW
the grid as shown in Fig. 12. It can be seen from Fig. 12 that the and reactive power of 4 kVAr to the loads as shown in Figs.
grid delivers 46 kW of real power and 4 kVAr of reactive power 15 and 16. At t = 0.2 s, the microgrid is disconnected from
to the loads. the distribution grid by the CBs due to a fault occurring in
At t = 0.2 s, PMSG A which generates real power of 5.5 kW the upstream network of the distribution grid. It can be seen
is connected to the dc grid. This causes a sudden power surge from Fig. 14 that the CBs fully separate the microgrid from the
at the dc grid and results in a voltage rise at t = 0.2 s as shown grid in about half a cycle, resulting in zero real and reactive
in the voltage waveform of Fig. 13. At t = 0.26 s, the EMS power supplied by the grid for 0.2 ≤ t < 0.4 s. With the loss of
increases the real delivered by each inverter to 10 kW while the power supply from the grid, the power imbalance between the
reactive power supplied by each inverter remains unchanged at generation and load demand is detected by the EMS. To maintain
4 kVAr as shown in Figs. 10 and 11. This causes a momentarily the stability of the microgrid, the SB is tasked by the EMS to
504 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 31, NO. 2, JUNE 2016

Fig. 14. Real (top) and reactive (bottom) power delivered by the grid. Fig. 16. Real (top) and reactive (bottom) power delivered by inverter 2.

Fig. 17. Real power delivered by SB.

Fig. 15. Real (top) and reactive (bottom) power delivered by inverter 1.

supply real power of 40 kW at t = 0.26 s as shown in Fig. 17. Fig. 18. DC grid voltage.
At the same time, the real and reactive power delivered by each
inverter is also increased by the EMS to 30 kW and 6 kVAr
as shown in Figs. 15 and 16 respectively. Fig. 18 shows the dc and reliability to the operation of the microgrid. However, the
grid voltage where slight voltage fluctuations are observed at proposed control design still requires further experimental val-
t = 0.26 s. The initial voltage rise at t = 0.26 s is due to the idation because measurement errors due to inaccuracies of the
power supplied by the SB while the subsequent voltage dip is voltage and current sensors, and modeling errors due to varia-
due to the increase in power drawn by the inverters. tions in actual system parameters such as distribution line and
transformer impedances will affect the performance of the con-
V. CONCLUSION troller in practical implementation. In addition, MPC relies on
In this paper, the design of a dc grid based wind power gen- the accuracy of model establishment, hence further research on
eration system in a microgrid that enables parallel operation of improving the controller robustness to modeling inaccuracy is
several WGs in a poultry farm has been presented. As compared required. The simulation results obtained and the analysis per-
to conventional wind power generation systems, the proposed formed in this paper serve as a basis for the design of a dc grid
microgrid architecture eliminates the need for voltage and fre- based wind power generation system in a microgrid.
quency synchronization, thus allowing the WGs to be switched
on or off with minimal disturbances to the microgrid opera- REFERENCES
tion. The design concept has been verified through various test [1] M. Czarick and J. Worley, “Wind turbines and tunnel fans,” Poultry Hous-
scenarios to demonstrate the operational capability of the pro- ing Tips, vol. 22, no. 7, pp. 1–2, Jun. 2010.
[2] The poultry guide: Environmentally control poultry farm ventilation sys-
posed microgrid and the simulation results has shown that the tems for broiler, layer, breeders and top suppliers. [Online]. Available:
proposed design concept is able to offer increased flexibility http://thepoultryguide.com/poultry-ventilation/
TAN et al.: CONTROL AND OPERATION OF A DC GRID-BASED WIND POWER GENERATION SYSTEM IN A MICROGRID 505

[3] Livestock and climate change. [Online]. Available: http://www. [27] T. Kato, K. Tamura, and T. Matsuyama, “Adaptive storage battery man-
worldwatch.org/files/pdf/Livestock%20and%20Climate%20Change.pdf. agement based on the energy on demand protocol,” in Proc. IEEE 3rd Int.
[4] Farm Energy: Energy efficient fans for poultry production. [Online]. Avail- Conf. Smart Grid Commun., 2012, pp. 43–48.
able: http://farmenergy.exnet.iastate.edu. [28] L. Xu and D. Chen, “Control and operation of a DC microgrid with variable
[5] A. Mogstad, M. Molinas, P. Olsen, and R. Nilsen, “A power conversion generation and energy storage,” IEEE Trans. Power Del., vol. 26, no. 4,
system for offshore wind parks,” in Proc. 34th IEEE Ind. Electron., 2008, pp. 2513–2522, Oct. 2011.
pp. 2106–2112. [29] C. A. Hill, M. C. Such, D. Chen, J. Gonzalez, and W. M. Grady, “Bat-
[6] A. Mogstad and M. Molinas, “Power collection and integration on the tery energy storage for enabling integration of distributed solar power
electric grid from offshore wind parks,” in Proc. Nordic Workshop Power generation,” IEEE Trans. Smart Grid, vol. 3, no. 2, pp. 850–857, Jun.
Ind. Electron., 2008, pp. 1–8. 2012.
[7] D. Jovic, “Offshore wind farm with a series multiterminal CSI HVDC,” [30] O. Gomis-Bellmunt, A. Junyent-Ferre, A. Sumper, and J. Bergas-Jane,
Elect. Power Syst. Res., vol. 78, no. 4, pp. 747–755, Apr. 2008. “Control of a wind farm based on synchronous generators with a cen-
[8] X. Lu, J. M. Guerrero, K. Sun, and J. C Vasquez “An improved droop con- tral HVDC-VSC converter,” IEEE Trans. Power Syst., vol. 26, no. 3,
trol method for DC microgrids based on low bandwidth communication pp. 1632–1640, Aug. 2011.
with DC bus voltage restoration and enhanced current sharing accuracy,” [31] W. Lu and B. T. Ooi, “Optimal acquisition and aggregation of offshore
IEEE Trans. Power Electron., vol. 29, no. 4, pp. 1800–1812, Apr. 2014. wind power by multiterminal voltage-source HVDC,” IEEE Trans. Power
[9] T. Dragicevi, J. M. Guerrero, and J. C Vasquez, “A distributed con- Del., vol. 18, no. 1, pp. 201–206, Jan. 2003.
trol strategy for coordination of an autonomous LVDC microgrid based [32] P. Cortes, G. Ortiz, J. I. Yuz, J. Rodriguez, S. Vazquez, and L. G. Franquelo,
on power-line signaling,” IEEE Trans. Ind. Electron., vol. 61, no. 7, “Model predictive control of an inverter with output LC filter for UPS
pp. 3313–3326, Jul. 2014. applications,” IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 1875–1883,
[10] N. L. Diaz, T. Dragicevi, J. C. Vasquez, and J. M. Guerrero, “Intel- Jun. 2009.
ligent distributed generation and storage units for DC microgrids—A [33] L. Wang, Model Predictive Control System Design Implementation Using
new concept on cooperative control without communications beyond MATLAB, vol. 1. New York, NY, USA: Springer, 2009.
droop control,” IEEE Trans. Smart Grid, vol. 5, no. 5, pp. 2476–2485, [34] C. Y. Teo, Principles Design Low Voltage Systems. Singapore: Byte Power
Sep. 2014. Publications, 1997.
[11] X. Liu, P. Wang, and P. C. Loh, “A hybrid AC/DC microgrid and its
coordination control,” IEEE Trans. Smart Grid, vol. 2, no. 2, pp. 278–286,
Jun. 2011.
[12] C. Jin, P. C. Loh, P. Wang, M. Yang, and F. Blaabjerg, “Autonomous
operation of hybrid AC-DC microgrids,” in Proc. IEEE Int. Conf. Sustain.
Energy Technol., 2010, pp. 1–7.
[13] J. A. P. Lopes, C. L. Moreira, and A. G.Madureira, “Defining control K. T. Tan (M’14) received the B.Eng. and Ph.D. degrees in electrical and elec-
strategies for microgrids islanded operation,” IEEE Trans. Power Syst., tronic engineering from Nanyang Technological University, Singapore, in 2008
vol. 21, no. 2, pp. 916–924, May 2006. and 2014, respectively. He then joined the School of Electrical and Electronic
[14] Y. Li, D. M. Vilathgamuwa, and P. Loh, “Design, analysis, and real-time Engineering, Nanyang Technological University, Singapore, as a Research Fel-
testing of a controller for multibus microgrid system,” IEEE Trans. Power low. He is currently a Lecturer in the Electrical Engineering Division, Ngee Ann
Electron., vol. 19, no. 5, pp. 1195–1204, Sep. 2004. Polytechnic, Singapore. His research interests are clean and renewable energy,
[15] C. L. Chen, Y. B. Wang, J. S. Lai, Y. S. Lai, and D. Martin, “Design microgrids, smart grids, and electric vehicles.
of parallel inverters for smooth mode transfer of microgrid applications,”
IEEE Trans. Ind. Electron., vol. 25, no. 1, pp. 6–15, Jan. 2010.
[16] S. Kouro, P. Cortes, R. Vargas, U. Ammann, and J. Rodrı́guez, “Model B. Sivaneasan (M’12) received the B.Eng. and Ph.D. degrees in electrical and
predictive control—A simple and powerful method to control power con- electronic engineering from Nanyang Technological University, Singapore, in
verters,” IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 1826–1838, Jun. 2007 and 2012, respectively. He is currently a Lecturer in the School of En-
2009. gineering, Nanyang Polytechnic, Singapore. Prior to his academic career, he
[17] K. S. Low and R. Cao, “Model predictive control of parallel-connected was a Research Fellow at Nanyang Technological University, Singapore. His
inverters for uninterruptible power supplies,” IEEE Trans. Ind. Electron., research interests are power line communications, advanced metering infrastruc-
vol. 55, no. 8, pp. 2884–2893, Aug. 2008. ture, electric vehicles, building energy management systems, demand response,
[18] J. Rodriguez and P. Cortes, Predictive Control Power Converters Electrical and smart grids.
Drives. Hoboken, NJ, USA: Wiley, 2012.
[19] S. Mariethoz, A. Fuchs, and M. Morari, “A VSC-HVDC decentralized
model predictive control scheme for fast power tracking,” IEEE Trans.
X. Y. Peng (S’11) received the B.Eng. degree in electrical and electronic engi-
Power Del., vol. 29, no. 1, pp. 462–471, Feb. 2014.
neering from Wuhan University, China, in 2010, and the M.Sc. degree in power
[20] M. Korpas and A. Holen, “Operation planning of hydrogen storage con-
engineering from Nanyang Technological University, Singapore, in 2011. He
nected to wind power operating in a power market,” IEEE Trans. Energy
is currently working toward the Ph.D. degree in the Clean Energy Research
Convers., vol. 21, no. 3, pp. 742–749, Sep. 2006.
Laboratory, School of Electrical and Electronic Engineering, Nanyang Techno-
[21] M. Falahi, S. Lotfifard, M. Eshani, and K. Butler-Purry “Dynamic model
logical University. His research interests include smart grids, control strategy in
predictive-based energy management of DG integrated distribution sys-
microgrids, and distributed energy resources.
tems,” IEEE Trans. Power Del., vol. 28, no. 4, pp. 2217–2227, Oct. 2013.
[22] D. E. Olivares, A. Mehrizi-Sani, A. H. Etemadi, C. A. Cañizares, R. Iravani
et al., “Trends in microgrid control,” IEEE Trans. Smart Grid, vol. 5,
no. 4, pp. 1905–1919, Jul. 2014. P. L. So (M’98–SM’03) received the B.Eng. degree with first class honors in
[23] N. Mendis, K. M. Muttaqi, S. Sayeef, and S. Perera, “Standalone op- electrical engineering from the University of Warwick, Coventry, U.K., in 1993,
eration of wind turbine-based variable speed generators with maximum and the Ph.D. degree in electrical power systems from Imperial College, Univer-
power extraction capability,” IEEE Trans. Energy Convers., vol. 27, no. 4, sity of London, London, U.K., in 1997. He is currently an Associate Professor
pp. 822–834, Dec. 2012. in the School of Electrical and Electronic Engineering, Nanyang Technological
[24] S. X. Chen, H. B. Gooi, and M. Q. Wang, “Sizing of energy storage for University, Singapore. Prior to his academic career, he worked for 11 years
microgrids,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 142–151, Aug. as a Second Engineer with China Light and Power Company Limited, Hong
2011. Kong, in the field of power system protection. His research interests include
[25] T. Dragicevic, J. M. Guerrero, J. C. Vasquez, and D. Skrlec, “Supervi- power system stability and control, power quality, power line communications,
sory control of an adaptive-droop regulated DC microgrid with battery clean and renewable energy, energy management, microgrids, smart grids, and
management capability,” IEEE Trans. Power Electron., vol. 29, no. 2, electric vehicles.
pp. 695–706, Feb. 2014. Dr. So was the Chair of the IEEE Singapore Section from 2009 to 2010. He
[26] M. Coleman, C. K. Lee, C. Zhu, and W. G. Hurley, “State-of-charge is currently a Member of the Electrical Testing Technical Committee, Singapore
determination from EMF voltage estimation: Using impedance, terminal Accreditation Council, and a Member of Working Group under the purview of
voltage, and current for lead-acid and lithium-ion batteries,” IEEE Trans. the Telecommunications Standards Technical Committee, Infocomm Develop-
Ind. Electron., vol. 54, no. 5, pp. 2550–2557, Oct. 2007. ment Authority of Singapore.