This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE SYSTEMS JOURNAL 1

Coordinated Control of Wind Turbine Blade Pitch

Angle and PHEVs Using MPCs for Load
Frequency Control of Microgrid
Jonglak Pahasa and Issarachai Ngamroo, Member, IEEE

AbstractThis paper proposes coordinated control of blade speed so that the generated power of wind turbine generator
pitch angle of wind turbine generators and plug-in hybrid electric (WTG) fluctuates and may cause a large frequency fluctuation
vehicles (PHEVs) for load frequency control of microgrid using problem [5][7].
model predictive controls (MPCs). The MPC is an effective model-
based predictive control, which calculates future control signals by In order to solve the frequency fluctuation problem, the
optimization method using plant model, current, and past signals control of WTG power generation of wind turbine is proposed
of the system. The MPC-based PHEVs power control can be used in [7][9]. Adjusting blade pitch angle of WTG for smoothing
to reduce frequency fluctuation of microgrid effectively. However, wind power output has been proposed in [10] and [11].
for large system, large number of PHEVs is needed to produce In addition, battery energy storage system (BESS) is a pow-
satisfying frequency deviation. In order to reduce the number
of PHEVs, the smoothing of wind power production by pitch erful tool for storing wind power output to maintain microgrid
angle control using MPC method is proposed and is coordinated frequency deviation [11], [12]. Coordinated control of wind and
with PHEVs control in this paper. The simulation results confirm battery is proposed in [3]. This work shows that controlling
that the coordinated control of pitch angle and PHEVs using WTG inverter is able to maximize wind energy. However, using
MPCs is able to reduce the number of PHEVs and the frequency BESS to suppress wind energy leads to high cost, particularly
fluctuation can be maintained significantly. In addition, the pro-
posed method is robust to the system parameters variation over when applied to a large system.
proportionalintegral derivative controllers. Smoothing wind power output in order to reduce the size of
battery has been proposed in [11]. This work shows that using
Index TermsBlade pitch angle, microgrid, model predictive
control (MPC), plug-in hybrid electric vehicle (PHEV), wind tur- the appropriate wind power reference, is able to reduce cost of
bine generator (WTG). BESS and loss of benefit from the wind energy.
On the other hand, plug-in hybrid electric vehicles (PHEVs)
are significantly expected to be installed in the customer side
I. I NTRODUCTION [5], [13][17]. With adequate energy stored in the battery of
PHEV, the bidirectional charging/discharging power control of
T HE microgrid concept has been gaining more attention
worldwide, particularly for the distribution system be-
cause microgrid is able to improve system reliability and
PHEV can be applied to alleviate frequency fluctuation [13]
[17]. Effective control of PHEVs can be employed for load
energy management capability [1], [2]. Microgrid consists of frequency control of microgrid with wind farm [5]. However,
multiple distributed generators (DGs) and associated loads [1]. for large system, large number of PHEVs is needed to produce
The penetration of renewable energy sources to the microgrid, satisfying frequency deviation.
particularly the wind power generation, tends to increase sig- In order to reduce both the number of PHEVs and fre-
nificantly because of low impact to environment and infinite quency fluctuation, smoothing of wind power production by
availability [2][4]. Nevertheless, wind speed is intermittent, pitch angle control and PHEV control using a model predictive
and the windmill output is proportional to the cube of wind control (MPC) method is proposed in this paper. The MPC
algorithm calculates future control signals each time instant for
tracking periodic reference signals [18][22]. The applications
of MPC to power systems have been successfully proposed
in the literature such as load frequency control [23], [24],
battery energy storage control [25], and PHEV control [26]. The
Manuscript received October 16, 2013; revised January 12, 2014 and main advantages of MPC over structured proportionalintegral
March 16, 2014; accepted March 24, 2014. This work was supported by
Thailand Research Fund under Grant MRG5680005.
derivative (PID) controllers are its ability to handle constraints,
J. Pahasa is with the Department of Electrical Engineering, School of robustness to model uncertainty, and noise [19].
Engineering, University of Phayao, Phayao 56000, Thailand (e-mail: jpahasa@ The organization of this paper is as follows. First, study
gmail.com).
I. Ngamroo is with the Department of Electrical Engineering, Faculty of system and modeling of microgrid system are explained in
Engineering, King Mongkuts Institute of Technology Ladkrabang, Bangkok Section II. Next, MPC design for blade pitch angle and PHEV
10520, Thailand (e-mail: [email protected]). controls are described in Section III. Subsequently, the experi-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. mental results are shown in Section IV. Finally, the conclusion
Digital Object Identifier 10.1109/JSYST.2014.2313810 is provided.

1932-8184 2014 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.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE SYSTEMS JOURNAL

Fig. 1. Microgrid system used in this paper.

Fig. 4. Pitch angle control law.

windmill, is the angular rotor speed for a windmill, and J is

the moment of inertia for a windmill.
The output power of generator Pg can be expressed as

3V 2 s(1 + s)R2
Pg = (4)
(R2 sR1 )2 + s2 (X1 + X2 )2

where s is the slip; V is the phase voltage; X1 and X2 are the

reactances of the stator and rotor, respectively; and R1 and R2
are the resistances of the stator and rotor, respectively. In the
Fig. 2. Linearized model of the microgrid system.
steady state, if energy loss is disregarded, Pw = Pg ; hence, Pw
can be approximated by

Pw = a1 () + a2 ()Vw2 ,
a1 () = 11 + 12 + 13 2 + 14 3 , (5)

a2 () = 21 + 22 + 23 2 + 24 3 ,

where ij is the constant. More details about the WTG system

Fig. 3. Windmill and generator. are provided in [10].
As in (5), Pw can be controlled by adjusting blade pitch angle
II. S TUDY S YSTEM AND M ODELING . Blade pitch angle control of WTG is appropriate adjustment
A. Microgrid System of the amount of wind received by the WTG to suppress the
output fluctuation by varying the pitch angle of WTG blade [8],
The microgrid system employed in this paper is shown in [10], [11]. In general, the pitch angle control is performed by
Fig. 1. System details are as follows: 20-MW thermal power, dividing the WTG power output curve into four control regions,
5.5-MW wind power, 3.5-MW PHEVs, and 15-MW load [5], as shown in Fig. 4, and control law for wind velocity Vw on each
[7][10]. The linearized model of microgrid system [5], [7] region is set as follows.
[10] is shown in Fig. 2. In this system, two MPCs are applied
to WTG blade pitch angle control and PHEV power control. Region 1: Vw Vwi ; set the pitch angle to 90 . In this region,
the WTG does not generate power.
Region 2: Vwi Vw Vwr , set the pitch angle = 10 , and
B. WTG System Pw is calculated by (5) so that the WTG can receive the
The block diagram of the WTG system [8], [10], [11] is maximum wind energy.
shown in Fig. 2. Windmill and generator models are shown Region 3: Vwr Vw Vwo , variably adjust the pitch angle
in Fig. 3. Windmill power output Pw can be calculated by the within the specified range (from 10 to 90 ), and Pw is
following equation: calculated by (5) so that the power output of WTG is
constant (rated power output) when is appropriately
Pw (Vw ) = 0.5Cp (, )Vw3 A (1) adjusted. Originally, the blade pitch angle control aims at
2 3 4
preventing the output of WTG from exceeding the rated
Cp (, ) = c1 () + c2 () + c3 () (2) capacity. The pitch angle conventionally can be changed
(, Vw ) = R/Vw (3) in region 3 actively. Here, the MPC-based real-time pitch
angle control is operated in region 3 for smoothing wind
where Vw is the wind speed, is the air density, A is the cross power output.
section of the rotor for a windmill, is the pitch angle, is the Region 4: Vwo Vw , set the pitch angle to 90 for safety, and
tip speed ratio, Cp is the power coefficient, R is the radius of a WTG does not generate power same as region 1.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PAHASA AND NGAMROO: CONTROL OF BLADE PITCH ANGLE AND PHEVs USING MPCs FOR MICROGRID CONTROL 3

Fig. 5. PHEV power control against frequency deviation.

Fig. 7. Basic concept of MPC.

III. MPC C ONTROLLER D ESIGN

A. MPC Method
Fig. 6. PHEV power control using MPC.
The MPC method is based on current measurements and
predictions of the future values of the outputs [18], [19]. The
Vw is the wind velocity, Vwi is the cut-in velocity, Vwr is objective of the MPC is to determine a sequence of control
the rated velocity, and Vwo is the cut-out velocity. In this paper, moves, i.e., manipulated input variable, so that the predicted
Vwi = 5, Vwr = 12, and Vwo = 24 m/s. response moves to the set point in an optimal manner. Fig. 7
shows the basic concept of MPC, where y is the actual output,
C. PHEV Power Control System y is the predicted output, and u is the manipulated input. At
the current sampling instant, denoted by k, the MPC strategy
PHEVs use bidirectional acdc converters that have the calculates a set of M values of the input {u(k + i 1), i =
capability to turn alternating current into direct current during 1, 2, 3, . . . , M }. The set consists of the current input u(k) and
battery charge mode and convert direct current into alternating M 1 future inputs. The input is held constant after the M
current in the battery discharge mode. Details of the power elec- control moves. The inputs are calculated so that a set of P
tronics topologies of bidirectional acdc and dcdc converters predicted outputs {y(k + i), i = 1, 2, . . . , P } reaches the set
for PHEV are provided in [27]. point in an optimal manner.
The PHEV power control against frequency deviation is The number of predictions P is referred to as the prediction
shown in Fig. 5 [12]. The V1G is the one-way charging power horizon, whereas the number of control moves M is called
control from grid to vehicles. The V2G is the bidirectional the control horizon. Although a sequence of M control moves
charging/discharging power control from vehicles to grid/from is calculated at each sampling instant, only the first move is
grid to vehicles. The PHEV power output of V2G (PV2G ) can actually implemented. Then, a new sequence is calculated at
be controlled as the droop characteristics against the frequency the next sampling instant; after new measurements become
deviation f as available again, only the first input move is implemented. This
 procedure is repeated at each sampling instant.
Kmax f, (|Kmax f | Pmax ) The MPC predictions are made using a dynamic model, typi-
PV2G = (6)
Pmax , (Pmax < Kmax f ) cally a linear empirical model such as a multivariable version of
the step response or difference equation models. Alternatively,
where Kmax and Pmax are the PHEV gain and the maximum
a transfer function or state-space models can be employed.
PHEV power, respectively. Kmax can be tuned by considering
The MPC method solves an optimization problem for finite
a tradeoff between the effect of vehicle and the fluctuation range
future time steps at current time [20]. Hence, the system can be
of state of charge according to the additional chargedischarge
represented by its finite impulse response [21] as
cycles. Here, Kmax of V2G is tuned by MPC based on de-
sired minimum frequency deviation. Pmax is defined by the
nT
200 V/25 A home outlet. y(k + 1) = y(k) + A i u(k i) (8)
In addition, V1G power PV1G is calculated by i=0

Kmax f, (0 < Kmax f Pmax ) where y(k) is the vector of manipulated moves at time instance
PV1G = Pmax , (Pmax < Kmax f ) (7) k, u(k) is the input at time instance k, nT is the number of
impulse response coefficients employed to model the system,
0, (Kmax f < 0).
A is the interaction matrix, and i is the coefficient number and
Fig. 6 shows PHEV power control using MPC. In the V2G can be defined as
model, the first-order transfer function with the time delay
TPHEV is used and is set as 1 [28] in this paper. i = gi+1 gi , i = 0, . . . , nT (9)
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE SYSTEMS JOURNAL

where gi is the scalar such that gi A is the ith impulse response In addition, the moving average may be used for smoothing
coefficient matrix. The MPC problem is to compute u(k) as the wind power output. However, the control of the pitch angle is
solution to the quadratic program (QP), which is defined as [21] not very easy because system behavior is both highly nonlinear
and quite uncertain [30]. MPC is more appropriate because the

M
smoothing wind power is adjusted by blade pith angle. The
min [y(k + j) r(k + j)]T Wy [y(k + j) r(k + j)]
u(k)M
j=1
MPC is used here for adjusting blade pith angle to produce
smoothing wind power output to reference value, based on wind
+ [u(k) u(k 1)]T Wu [u(k) u(k 1)]
speed and feedback control signal of wind power generation.
nT
Subject to y(k + 1) = y(k) + A i u(k i) Since the blade pitch angle is controlled through the hy-
i=0 draulic servo system, there are constraints on the blade pitch
umax + u(k 1) u(k) umax + u(k 1) (10) angle , and its rate of change d /dt can be expressed as

where r(k + j) is the desired profile, Wy and Wu are the 10[deg] 30[deg] (11)
positive semidefinite weighting matrices, and M is the control |d/dt| 10[deg /s]. (12)
horizon. Each weight (Wy , Wu ) is assumed to be a constant
multiplied by the identity matrix, which is appropriate for WTG For MPC method, the WTG can be represented by its finite
blade pitch angle and PHEV control. In particular, Wu is often impulse response as
selected large enough that rate constraints are satisfied.
nT
Although, MPC allows design of interactive and multi- PWTG (k + 1) = PWTG (k) + AWTG i (k i) (13)
inputmulti-output systems [20]. However, in this paper, the i=0
properties of pitch angle and PHEV controls are different. Pitch
angle control objective is to smooth wind power output with where (k) is the vector of manipulated moves (blade pitch
power generation and set point as manipulate and reference angle) at time instance k, PWTG (k) is the WTG power gener-
signals. In contrast, PHEV control objective is to reduce system ation at time instance k, nT is the number of impulse response
frequency fluctuation with system frequency and reference coefficients employed to model the system, and AWTG is the
frequency as manipulate and reference signals. Here, PHEV interaction matrix of WTG. In this paper, since the windmill
control is more complicated and needs more precise control and generator of the WTG system are represented by a mathe-
action than pitch angle control. Pitch angle control needs faster matical model as in (5), the mapping from inputs to outputs is
response to wind speed, whereas PHEV control responds to assumed to be linear over the operating region, which is a good
system frequency deviation. assumption in practice so that AWTG can be defined as 1 [22]
Moreover, the smoothing wind power controlled by MPC- for WTG system.
based pitch angle control is resulted to frequency deviation.
However, pitch angle control depends on WTG power genera- C. MPC for PHEV Control
tion and wind speed, which are not related to part of MPC-based
PHEV control. In addition, the control of pitch angle and some The control loop of PHEV using MPC2 controller is shown
part of WTG are modeled by mathematic equation, whereas in Fig. 2. The state space of the PHEV control loop can be
PHEV is modeled by first-order transfer function with time lag defined as




TPHEV . Consequently, the two single MPCs are appropriately PPHEV 1/TPHEV 0 PPHEV
employed to control pitch angle and PHEV power output in this =
f 1/M D/M f
paper.

1/TPHEV
In addition, two simple PI or fuzzy controllers may be + uMPC (14)
0
sufficient for pitch angle and PHEV output control. However,

using a PI controller is not adequate when pitch angle is held PPHEV
[y] = [0 1] + [0]UMPC (15)
constant, which may occur when wind speed is lower than set f
point for a long time. In addition, when using a fuzzy controller, where PPHEV is the charging power of PHEV, f is the
it is not easy to set a fuzzy rule for a specific problem [29]. change of system frequency, uMPC is the change of control
signal produced by MPC, TPHEV is the time constant of PHEV,
B. MPC for WTG Blade Pitch Angle Control and M and D are the inertia constant and damping coefficient
of the microgrid system, respectively. This state space is used
Control loop of WTG blade pitch angle control using MPC1 as model in the MPC calculations. Therefore, the MPC method
is shown in Fig. 2. Input of MPC consists of the difference applied to PHEV control can be defined as
of WTG output power PWTG and reference PWTG,ref . Blade
pitch angle is controlled using these inputs. Here, generator x(k + 1) = APHEV x(k) + BPHEV u(k) (16)
torque can be used as a crucial feedback control signal in y(k) = CPHEV x(k) + DPHEV u(k) (17)
maximizing the output power of WTG and can be used as input


for smoothing wind power output. Nevertheless, using power 1/TPHEV 0 1/TPHEV
where APHEV = , BPHEV = ,
generation as a feedback control signal is more adequate for 1/M D/M 0
smoothing wind power output [11]. CPHEV = [ 0 1 ], and D = [0].
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PAHASA AND NGAMROO: CONTROL OF BLADE PITCH ANGLE AND PHEVs USING MPCs FOR MICROGRID CONTROL 5

Fig. 8. Convergence curve for MPC weight tuning.

Fig. 9. Random load pattern.

D. MPC Weight Tuning

The MPC weights (i.e., Wy and Wu ) for pitch angle and
PHEV controls are simultaneously tuned by particle swarm
optimization (PSO) [31] using the integral absolute error (IAE)
[5] as the objective function as follows:
 
Minimize IAEMPC = |Pg_ref Pg |dt + |f |dt
0 0 Fig. 10. Wind speed of case studies.
Subject to Wyi, min Wyi Wyi, max
Wui, min Wui Wui, max , i = 1, 2 (18)

where Wyi is the input weight of the ith MPC; Wui is the output
weight of the ith MPC; Wyi,min and Wyi,max are the minimum
and maximum input weights of the ith MPC, respectively;
Wui,min and Wui,max are the minimum and maximum output
weights of the ith MPC, respectively; and i = 1, 2 for MPC1
(pitch angle control) and MPC2 (PHEV control), respectively.

IV. S IMULATION R ESULTS

Computer simulations have been carried out in order to
validate the effectiveness of the proposed scheme. A system
shown in Fig. 2 is used as the study system.
In the design of MPC, the parameters of the MPC controller
for WTG blade pitch angle control are set as follows: prediction
horizon P = 10, control horizon M = 3, and sampling interval
S = 0.1 s. In addition, the parameters of the MPC controller for
PHEVs are set as follows: P = 10, M = 3, and S = 0.1 s.
For the optimization of MPC weights, the convergence curve
of optimization is shown in Fig. 8; the optimal weights are
as follows: Wy1 = 0.6780, Wu1 = 0.6312, Wy2 = 0.0850, and
Wu2 = 1.8653.
In the simulation studies, it is supposed that the studied Fig. 11. Absolute maximum frequency deviation when the number of PHEVs
microgrid is performed under the random load pattern and wind changed: (a) case 1; (b) case 2; and (c) case 3.
speed of the three case studies, as shown in Figs. 9 and 10,
respectively. The detailed models of random load pattern and Case 3: Wind speed is set at 1025 m/s. WTG operates in
wind speed are provided in Appendixes A and B, respectively. regions 2, 3, and 4. The effects of wind speed working
Details about wind speed and WTG generation of the case between cut-in velocity, cutoff velocity, and MPC control
studies are as follows. region are investigated.
Case 1: It is assumed that wind speed is 024 m/s. WTG works In addition, the proposed coordinated control of WTG
in regions 1, 2, and 3 (see Fig. 4). In this case, WTG blade pitch angle and PHEVs using MPCs called MPC-
normally operates in region 2. However, MPC is activated Pitch-PHEVs is compared with the MPC-Pitch and MPC-
in region 3. Hence, the effect of short-period control of PHEVs. The MPC-Pitch is the system with MPC control wind
MPC is investigated in this case. power output via blade pitch angle and without PHEVs in the
Case 2: It is assumed that wind speed is 1024 m/s. WTG system. The MPC-PHEVs is the system with MPC for PHEV
works in regions 2 and 3. In this case, WTG normally power control and without smoothing wind power output.
operates in region 3. Thus, the effect of long period of MPC Figs. 11(a)(c) shows maximum frequency deviation when the
control in region 3 is investigated. number of PHEVs is changed for cases 1, 2, and 3, respectively.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE SYSTEMS JOURNAL

In Fig. 11(a), the maximum frequency deviation of the MPC-

Pitch is constant, whereas MPC-PHEVs and the proposed coor-
dinated control methods are able to reduce maximum frequency
deviation when the number of PHEVs increases. Note that
when the number of vehicles = 0, it is implied that the PHEVs
are not included in the microgrid system.
In Fig. 11(b), the maximum frequency deviation in the case
of MPC-PHEVs is higher than in the case of MPC-Pitch
and MPC-Pitch-PHEVs when the number of PHEVs is small.
When the number of PHEVs increases, the maximum frequency
deviation in the case of MPC-PHEVs decreases. However,
the maximum frequency deviation in the case of MPC-Pitch-
PHEVs is better than that of MPC-Pitch when the number of
PHEVs is increased.
In Fig. 11(c), the maximum frequency deviation of the
proposed coordinated control methods can be reduced using
small number of PHEVs. The maximum frequency deviation
in the case of MPC-PHEVs is slightly reduced when the
number of PHEVs is much increased, whereas MPC-Pitch is
constant.
Note that the number of PHEVs has little effects to the MPC
design and does not increase the order of the system. From
the experiment, the design of one MPC model can be used
for various numbers of PHEVs. However, in order to produce Fig. 12. Time simulation results of case 1: (a) frequency deviation; (b) wind
the best performance, each number of PHEVs uses its MPC power output; and (c) PHEV power output.
model.
For the time simulation of the three case studies, 4 105 ve-
hicles (i.e., 0.0015 MW each) are employed. For cases 1 and 2,
it is assumed that PHEVs are not connected at the beginning
of simulation time. Then, PHEVs are randomly connected at
30 min (0.5 h) to suppress frequency fluctuations. In case 3,
it is assumed that the PHEVs are randomly connected to the
microgrid system at 0.5 h, disconnected at 2.0 h, and then
reconnected at 2.5 h.
Figs. 12(a)(c) shows frequency deviation and the power out-
puts of WTG and PHEVs of case 1, respectively. The frequency
deviation in the case of MPC-PHEVs and MPC-Pitch-PHEVs
is clearly better than that in the case of MPC-Pitch.
Fig. 13 shows the time simulation results of case 2. In
Fig. 13(a), at the beginning of simulation time, the frequency
deviation in the case of MPC-Pitch-PHEVs is similar to MPC-
Pitch and better than MPC-PHEVs because the effect of MPC-
based smoothing wind power output. After PHEVs are applied
to the system at 0.5 h, MPC-Pitch-PHEVs are able to control
the frequency of the system effectively. In contrast, PMC-
Pitch and MPC-PHEVs cannot control the frequency of the
system. Here, MPC-PHEVs work not well because PHEVs
charging/discharging power values are not enough to maintain
power fluctuation. On the contrary, the MPC-Pitch-PHEVs Fig. 13. Time simulation results of case 2: (a) frequency deviation; (b) wind
are capable to control the frequency for all simulation time power output; and (c) PHEV power output.
effectively because smoothing wind power using MPC-based
pitch angle control can reduce effects of wind power output and Figs. 14(a)(c) shows frequency deviation, wind power out-
can help PHEVs to use low charging/discharging power. put, and PHEVs power output of case 3, respectively. The
In addition, when wind speed is lower than rated velocity, MPC-Pitch and MPC-PHEVs cannot control the frequency,
as shown in Fig. 13(b), the system frequency in the case of whereas the MPC-Pitch-PHEVs are able to control the fre-
MPC-Pitch-PHEVs can be maintained in acceptable ranges. quency successfully.
This result implies that the MPC-Pitch-PHEVs are robust to Fig. 15 shows the blade pitch angle of the case studies.
wind power variation. Pitch angle varies between 10 and 30 and has not fluctuated
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PAHASA AND NGAMROO: CONTROL OF BLADE PITCH ANGLE AND PHEVs USING MPCs FOR MICROGRID CONTROL 7

Fig. 14. Time simulation results of case 3: (a) frequency deviation; (b) wind Fig. 16. Frequency deviation: (a) case 1; (b) case 2; and (c) case 3.
power output; and (c) PHEV power output.
wind power generation deviates from the set point. As a result,
the optimal PID controllers are obtained as follows:
PID-WTG : Kp = 0.5034, Ki = 0.0027, Kd = 0
PID-PHEV : Kp = 0.2644, Ki = 0.0065, Kd = 0
where Kp , Ki , and Kd are the proportional, integral, and
derivative gains of the PID controller, respectively.
Fig. 16 shows the comparison of frequency deviation be-
tween the proposed MPC and PID. The MPC is able to produce
better frequency deviation than the PID controller.
In order to confirm the robustness of the control method
against system parameter uncertainty and the absolute maxi-
mum frequency deviation when system parameters (i.e., M and
D) are changed from nominal operating condition to 50%, M
is changed from 9.02 to 4.52 and D is changed from 2 to 1 [5],
[28], as shown in Fig. 17. Clearly, the MPC is robust to system
parameter variation over the PID controller.
The simulation results of the three case studies clearly con-
firm that the coordinated control of the MPC-Pitch-PHEVs is
better than the MPC-Pitch and MPC-PHEVs. The frequency
fluctuation of the system can be reduced by the MPC-PHEVs.
In addition, the number of PHEVs can be reduced by smooth-
ing wind power output-based MPC blade pitch angle control.
Fig. 15. Blade pitch angle of the case studies: (a) case 1; (b) case 2; and
Therefore, using both the MPC-Pitch and MPC-PHEV methods
(c) case 3. not only reduces frequency fluctuation but also reduces the
number of PHEVs. Additionally, the MPC is robust to system
much during the simulation time, which are consistent with the parameter variation over PID controller.
physically limited WTG provided in [30].
Additionally, in order to demonstrate the effectiveness of the V. C ONCLUSION
proposed method, comparison of the MPC and PID controller Coordinated control of blade pitch angle and PHEV power
is investigated. The PID parameters are tuned by PSO [31]. output using MPCs in order to reduce fluctuation of frequency
The objective function of the optimization is defined based on in microgrid system has been proposed in this paper. The MPC
minimization of the IAE [5] of the frequency deviation, and for blade pitch angle control is employed for smoothing wind
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE SYSTEMS JOURNAL

Fig. 19. Wind speed model.

A PPENDIX
W IND S PEED M ODEL
The wind speed model [32], [33] is shown in Fig. 19. The
wind speed standard deviation is multiplied by a random output
fluctuation derived from the white noise block with a low-pass
filter in MATLAB/SIMULINK in order to evaluate the random
wind speed fluctuation.

R EFERENCES
[1] E. Alegria, T. Brown, E. Minear, and R. H. Lasseter, CERTS microgrid
demonstration with large-scale energy storage and renewable generation,
IEEE Trans. Smart Grid, vol. 5, no. 2, pp. 937943, Mar. 2014.
[2] B. Kroposki, R. Lasseter, T. Ise, S. Morozumi, S. Papathanassiou, and
N. Hatziargyriou, Making microgrids work, IEEE Power Energy Mag.,
vol. 6, no. 3, pp. 4053, May/Jun. 2008.
Fig. 17. Absolute maximum frequency deviation when system parameters [3] J. Han, S. K. Solanki, and J. Solanki, Coordinated predictive control
change: (a) case 1; (b) case 2; and (c) case 3. of a wind/battery microgrid system, IEEE Emerging Sel. Topics Power
Electron., vol. 1, no. 4, pp. 296305, Dec. 2013.
[4] P.-K. Keung, P. Li, H. Banakar, and B. T. Ooi, Kinetic energy of wind-
turbine generators for system frequency support, IEEE Trans. Power
Syst., vol. 24, no. 1, pp. 279287, Feb. 2009.
[5] S. Vachirasricirikul and I. Ngamroo, Robust controller design of heat
pump and plug-in hybrid electric vehicle for frequency control in a smart
microgrid based on specified-structure mixed H2 / H control technique,
Appl. Energy, vol. 88, no. 11, pp. 38603868, Nov. 2011.
[6] T. Masuta and A. Yokoyama, Supplementary load frequency control by
use of a number of both electric vehicles and heat pump water heaters,
IEEE Trans. Smart Grid, vol. 3, no. 3, pp. 12531262, Sep. 2012.
Fig. 18. Load model. [7] T. H. Mohamed, J. Morel, H. Bevrani, and T. Hiyama, Model predictive
based load frequency control design concerning wind turbines, Elect.
power production of WTG. In addition, the MPC for PHEVs Power Energy Syst., vol. 43, pp. 859867, 2012.
[8] T. Senjyu, T. Kaneko, A. Uehara, A. Yona, H. Sekine, and C.-H. Kim,
controller is employed in order to control load frequency of the Output power control for large wind power penetration in small power
microgrid system. Simulation results performed in the studied system, Renew. Energy, vol. 34, no. 11, pp. 23342343, Nov. 2009.
microgrid system show that the proposed coordinated control [9] H. Bevrani and P. R. Daneshmand, Fuzzy logic-based load-frequency
control concerning high penetration of wind turbines, IEEE Syst. J.,
of MPC-Pitch-PHEVs has better performance than MPC-Pitch vol. 6, no. 1, pp. 173180, Mar. 2012.
and MPC-PHEVs. The results imply that the proposed MPC- [10] T. Senjyu, R. Sakamoto, N. Urasaki, T. Funabashi, H. Fujita, and
based coordinated control method not only reduces frequency H. Sekine, Output power leveling of wind turbine generator for all
operating regions by pitch angle control, IEEE Trans. Energy Convers.,
fluctuation of the system but also reduces the number of vol. 21, no. 2, pp. 467475, Jun. 2006.
PHEVs. Moreover, the proposed MPC-based pitch angle and [11] Y. Nishizaki, H. Irie, A. Yokoyama, and Y. Tada, Blade pitch angle
PHEV control is robust to the system parameter variation when control and its capacity reduction effect on battery for load frequency
control in power system with a large capacity of wind power generation,
compared with PID controller. IEEJ Trans. Power Energy, vol. 129, no. 1, pp. 5056, 2009.
[12] Y. Ota, H. Taniguchi, T. Nakajima, K. M. Liyanage, K. Shimizu,
A PPENDIX T. Masuta, J. Baba, and A. Yokoyama, Effect of smart storage in ubiq-
L OAD M ODEL uitous power grid on frequency control, IEEJ Trans. Power Energy,
vol. 131, no. 1, pp. 94100, 2011.
Modeling of system load is shown in Fig. 18. The input [13] J. M. Foster, G. Trevino, M. Kuss, and M. C. Caramanis, Plug-in electric
vehicle and voltage support for distributed solar: Theory and application,
power variability of microgrid system load is determined by IEEE Syst. J., vol. 7, no. 4, pp. 881888, Dec. 2013.
considering the deviation from the initial value. In addition, [14] M. Takagi, H. Yamamoto, K. Yamaji, K. Okano, R. Hiwatari, and T. Ikeya,
the standard deviation d PLoad as in (A1) is multiplied by the Load frequency control method by charge control for plug-in hybrid
electric vehicles with LFC signal, IEEJ Trans. Power Energy, vol. 129,
random output fluctuation derived from the white noise block no. 11, pp. 13421348, 2009.
in MATLAB/SIMULINK in order to simulate the real-time [15] Y. Ota, H. Taniguchi, T. Nakajima, K. M. Liyanage, J. Baba, and
random power fluctuation on the load side. The deviation for A. Yokoyama, Autonomous distributed V2G (vehicle-to-grid) satisfying
scheduled charging, IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 559564,
the system load is simulated close to an actual change wave by Mar. 2012.
the following functions [32], [33]: [16] C.-T. Li, C. Ahn, H. Peng, and J. Sun, Synergistic control of plug-in
vehicle charging and wind power scheduling, IEEE Trans. Power Syst.,
dPLoad = 0.6 PLoad . (A1) vol. 28, no. 2, pp. 11131121, May 2013.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

PAHASA AND NGAMROO: CONTROL OF BLADE PITCH ANGLE AND PHEVs USING MPCs FOR MICROGRID CONTROL 9

[17] H. Liu, Z. Hu, Y. Song, and J. Lin, Decentralized vehicle-to-grid control [31] J. Kennedy and R. Eberhart, Particle swarm optimization, in Proc. IEEE
for primary frequency regulation considering charging demands, IEEE Int. Conf. Neural Net., 1995, vol. 4, pp. 19421948.
Trans. Power Syst., vol. 28, no. 3, pp. 34803489, Aug. 2013. [32] X. Li, D. Hui, X. Lai, and T. Yan, Power quality control in wind/fuel
[18] D. E. Seborg, T. F. Edgar, and D. A. Mellichamp, Process Dynamics and cell/battery/hydrogen electrolyzer hybrid micro-grid power system, in
Control, 2nd ed. Hoboken, NJ, USA: Wiley, 2003. Applications and Experiences of Quality Control. Rijeka, Croatia: In-
[19] K. S. Holkar and L. M. Wanghmare, An overview of model predictive Tech, 2011, pp. 579594.
control, Int. J. Control Autom., vol. 3, no. 4, pp. 4764, Dec. 2010. [33] T. Michigami and T. Ishii, Construction of fluctuation load model and
[20] M. G. Na, D. W. Jung, S. H. Shin, J. W. Jang, K. B. Lee, and Y. J. Lee, dynamic simulation with LFC control of DC power system and frequency
A model predictive controller for load-following operation of PWR reac- converter interconnection, in Proc. IEEE/PES Transmiss. Distrib. Conf.
tors, IEEE Trans. Nucl. Sci., vol. 52, no. 4, pp. 10091020, Aug. 2005. Exhib., Asia-Pacific, 2002, pp. 382387.
[21] J. G. VanAntwerp and R. D. Braatz, Fast model predictive control of
sheet and film processes, IEEE Trans. Control Syst. Technol., vol. 8,
no. 3, pp. 408417, May. 2000.
[22] M. Morari and N. L. Ricker, Model Predictive Control Toolbox. Natick,
MA, USA: MATH WORKS Inc., 1998.
[23] T. H. Mohamed, H. Bevrani, A. A. Hassan, and T. Hiyama, Decentral- Jonglak Pahasa received the B.Eng. degree in
ized model predictive based load frequency control in an interconnected electrical engineering from King Mongkuts Insti-
power system, Energy Convers. Manage., vol. 52, no. 2, pp. 12081214, tute of Technology Ladkrabang (KMITL), Bangkok,
Feb. 2011. Thailand, in 1997; the M.Eng. degree in electrical
[24] M. Khalid and A. V. Savkin, An optimal operation of wind energy storage engineering from Chiang Mai University, Chiang
system for frequency control based on model predictive control, Renew. Mai, Thailand, in 2007; and the D.Eng. degree in
Energy, vol. 48, pp. 127132, Dec. 2012. electrical engineering from KMITL in 2011.
[25] A. Khatamianfar, M. Khalid, A. V. Savkin, and V. G. Agelidis, Improving Since July 2010, she has been with the School
wind farm dispatch in the Australian electricity market with battery energy of Engineering, University of Phayao, Phayao,
storage using model predictive control, IEEE Trans. Sustain. Energy, Thailand. Her fields of interest are the application of
vol. 4, no. 3, pp. 745755, Jul. 2013. artificial intelligence in power system stability and
[26] F. Yan, J. Wang, and K. Huang, Hybrid electric vehicle model predictive control.
control torque-split strategy incorporating engine transient characteris-
tics, IEEE Trans. Veh. Tech., vol. 61, no. 6, pp. 24582467, Jul. 2012.
[27] J. Gallardo-Lozano, M. I. Milans-Montero, M. A. Guerrero-Martnez,
and E. Romero-Cadaval, Electric vehicle battery charger for smart grids,
Electr. Pow. Syst. Res., vol. 90, pp. 1829, Sep. 2012.
[28] S. Vachirasricirikul and I. Ngamroo, Robust LFC in a smart grid with Issarachai Ngamroo (M12) received the B.Eng.
wind power penetration by coordinated V2G control and frequency con- degree from King Mongkuts Institute of Technology
troller, IEEE Trans. Smart Grid, vol. 5, no. 1, pp. 371380, Jan. 2014. Ladkrabang (KMITL), Bangkok, Thailand, in 1992
[29] T. Shibata, T. Fukuda, and K. Tanie, Synthesis of fuzzy, artificial intel- and the M.Eng. and Ph.D. degrees from Osaka Uni-
ligence, neural networks, and genetic algorithm for hierarchical intelli- versity, Suita, Japan, in 1997 and 2000, respectively,
gent control-top-down and bottom-up hybrid method, in Proc. IJCNN, all in electrical engineering.
Nagoya, Japan, 1993, vol. 3, pp. 28692872. Currently, he is an Associate Professor with the
[30] H. Geng and G. Yang, Robust pitch controller for output power levelling Department of Electrical Engineering, Faculty of
of variable-speed variable-pitch wind turbine generator systems, IET Engineering, KMITL. His research interests include
Renew. Power Gener., vol. 3, no. 2, pp. 168179, Jun. 2009. power system stability, dynamics, and control.