Proceedings of the 37th Chinese Control Conference

July 25-27, 2018, Wuhan, China

Model predictive load frequency control of isolated micro-grid

with electrical vehicles
Junjun Pan, Shaoyuan Yu, Miaomiao Ma∗
School of Control & Computer Engineering, North China Electrical Power University, Beijing 102206
E-mail: [email protected]

Abstract: Electrical vehicles (EVs) can be used as mobile energy storage units, which could be an effective solution for LFC in
an isolated micro-grid. This paper design the load frequency controller based on model predictive control (MPC) for micro-grid
with electrical vehicles. The MPC controller is designed by minimizing the objective function. Meanwhile, the limitations of the
output power increment and the output power increment rate of each subsystem are also considered in this paper. The simulation
results show that when the system is subjected to a load step fluctuation, the proposed MPC can reasonably allocate the output of
each subsystem, and quickly eliminate the frequency fluctuation, while respecting the equivalent EVs and diesel power system
constraints.

Key Words: Load frequency control, Model predictive control, Constraints, Micro-grid, Electrical vehicle

1 Introduction can act as mobile storage units, which could be a solution for
load frequency control (LFC) in an isolated mirco-grid sys-
With the development of economy, demand for electric-
tem. In this paper, we focus on the design of a power system
ity has also grown rapidly. The micro-grid has become an
frequency oscillation damping controller using EVs.
alternate choice to compete with traditional centralized elec-
In power systems, the responsibility of the LFC is to create
tricity plants for its small-scale and on-site distributed ener-
a power balance in the system so that the system frequen-
gy providing capability [1]. A typical micro-grid system in-
cy changes only within certain bounds and the reasonably
cludes distributed generation units, such as diesel generators,
dynamic performance of the system is sustained. Recently,
renewable sources and loads. A micro-grid can be operated
more and more researchers pay attention to the LFC problem
in isolated mode or grid-connected mode [2]. If a micro-grid
of micro-grid system and many different control methods
is grid-connected, the loads can be supplied by the connect-
have been suggested in order to achieve better control per-
ed electrical power system. Otherwise, the distributed gen-
formance, based on various control techniques such as fuzzy
eration has to cope with varying load in the isolated mode,
control (e.g. [6–9]), sliding-mode control (e.g. [10, 11]),
which requires proper control of micro-grid to ensure fre-
adaptive control (e.g. [12, 13]) and model predictive con-
quency and voltage stability in order to supply high-quality
trol (e.g. [14–17]). In [7], the author proposes a combina-
to customers.
tion of the general type-2 fuzzy logic sets (GT2FLS) and
With the development of the vehicle-to-grid (V2G) tech-
the Modified Harmony Search Algorithm (MHSA) tech-
nique, electrical vehicles (EVs) are considered to be an in-
nique, as a novel heuristic algorithm, to adaptively tune the
evitable trend in the automotive industry [3]. It is well
proportional-integral (PI) controller for LFC in island micro-
known that EV is not only energy efficient, but also environ-
grids. The proposed control approach is easy to implemen-
mentally friendly. What’s more, EV technology is a promis-
t and it can handle both structured and unstructured uncer-
ing approach to reduce gasoline consumption, to eliminate
tainties. However, it can only demonstrated good dynamic-
smog and to reduce the emission of greenhouse gas. When
s when some specific membership functions were selected,
EVs are connected to a micro-grid, they can be a perfect so-
whose application can be limited in reality. In [16], the au-
lution for the load disturbance, i.e., a charging EV can be
thor proposes a new coordinated LFC controller based on
regarded as a load for the grid [4], while a discharging EV
multivariable generalized predictive control (MGPC). With
can be regarded as a power source for the grid [5]. Statis-
the control procedure of multi-step prediction, rolling op-
tics show that private cars stay idle for 95% of a day, which
timization and feedback correction [18], MGPC can solve
means most electrical vehicles are still connected to the grid
multivariable optimal control problem with constraints. But,
in charge or idle mode. By applying V2G technology, EVs
the shortcoming of the proposed control approach is its
This work is financially supported by the Fundamental Research Funds heavy burden of computations. Due to the fact that when
for the Central Universities (No.2017ZZD004) and the Beijing Natural Sci- MGPC is applied to practical engineering, it’s too sophis-
ence Foundation (No. 4173079). ticated, and there are many parameters need to be adjust-
∗ Corresponding author: [email protected]

3588
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.
 ed. Therefore, it increases the complexity of practical en- this paper considers an equivalent EV model to deal with
gineering applications. Moreover, it has a large amount of the different inverter capacities of different kinds of EVs.
computation and long online computing time, which is not Fig. 2 shows the equivalent EV model for load frequency
suitable for real-time control system requiring fast response. control [6]. This model is used to simulate the charging and
What’s more, most classical control (e.g. [7, 13]) methods discharging behavior of the EV battery.
mentioned above could not achieve satisfactory performance
since the effects of the output power increment and the out- PH PH
. !  . ! 
put power increment rate constraint were not explicitly con- GH
'X (
   PH
'3(
sidered in the controller design and only imposed on the sys- 7HV   PH 
 GH
tems during simulation.  PH .    PH .  
. .
In order to overcome the aforementioned problem, we ( PD[ ( PLQ

propose the MPC scheme for the LFC problem in the micro- (
grid system with electrical vehicles. The novel contribution 7RWDOHQHUJ\PRGHO

of this paper is the application of MPC to the LFC of the

micro-grid system, considering output power increment and Fig. 2: The equivalent EV model for LFC.
the output power increment rate constraints, and also condi-
tions of loads disturbance. The results confirm the superior- As shown in the figure, Te represents the time constant of
ity of the proposed MPC technique. EV, ΔuE is the LFC signal sent to EV, ±μe , ±δe represen-
The paper is organized as follows. Section 2 describes the t the inverter capacity and the power ramp rate constraints,
model of the micro-grid power system with electrical vehi- respectively. Also, E denotes the energy of the EV battery
cles to be studied. In Section 3, we state briefly the design of at present time. Emax and Emin denote the maximum and
MPC controller for micro-grid power system with electrical minimum controllable energy of the EV battery. ΔPE rep-
vehicles. Both simulation and analysis results are given and resents the charging or discharging power of the EV. When
discussed in Section 4. Finally, the conclusion is given in ΔPE = 0, the EV is in the idle state. In this case, the EV
Section 5. neither charges nor discharges. When ΔPE > 0, it means
2 Load frequency control model of an isolated that the EV is in the discharging state. Finally, ΔPE < 0
micro-grid with electrical vehicles means EV is in the charging state. It should be noted that the
EV can only charge and discharge within the range of ±μe .
Fig. 1 shows the framework of the proposed LFC con- The dynamic of the equivalent EV model can be expressed
troller in an isolated micro-gird, which consists of a diesel as:
generator (DG), two equivalent EVs, and load disturbance 1 1
ΔṖE = ΔuE − ΔPE (1)
ΔPD . Also, in the figure, 2Ht represents the equivalent in- Te Te
ertia constant of the isolated micro-grid, which consists of In the actual system, the charging and discharging process
inertias for all the directly connected generators and motor of EVs should satisfy the following constraints:
loads.
− μe  ΔPE (t)  μe (2a)


− δej  ΔṖEj (t)  δej , j = 1, 2 (2b)
5

'X '*   2.2 Model of the diesel power system

'* /RDG

'3'* '3' Generally, diesel generator (DG) is a small scale power

/)& 'X (    generation equipment with the characteristics of fast start-
 'I
&RQWUROOHU (9  + WV
'3( 
  ing speed, durability and high efficiency. DG can use pow-
'X (  '3(  er control mechanisms to follow the load demand variation-
(9
s in a short time. when the power demand fluctuates, DG
Fig. 1: The model of the micro-grid with electrical vehicles changes its output through fuel regulation. The continuous
including LFC controller time transfer function model of the DG for LFC is shown in
Fig. 3. It’s clearly seen from the figure that the DG model
consists of a governor and a generator, which are represented
2.1 Model of the electrical vehicle by first order inertial devices. The model of DG represents
In recent years, EVs have played an important role in our the relationship between the control signal and the output
life. And there are many kinds of EVs in each EV station, power. In Fig. 3, Δf is the frequency deviation, ΔuDG is

3589
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.
 
'I x = Δf ΔPDG ΔXG ΔPE1 ΔPE2

5 'LHVHO  T
*RYHUQRU
*HQHUDWRU
GGJ PGJ u = ΔuDG ΔuE1 ΔuE2
'X '* 
'[ * 

7J V   7G V  
'3'* d = ΔPD
 GGJ  PGJ 
y = Δf

Fig. 3: The transfer function model of DG for LFC And, x ∈ R5 is the state vector, u ∈ R3 is the controlled
input, d ∈ R is the uncontrolled input (power disturbance),
y ∈ R is the system output, correspondingly.
the LFC signal sent to DG, ΔXG is the governor valve po-
sition increment. Tg , Td represent the time constant of the
3 Design of model predictive controller
governor and the DG, respectively. Also, R stands for the Now we design the the load frequency controller based
speed regulation coefficient of the DG, ±δdg are the limits on model predictive control for the micro-grid with electri-
of the power ramp rate and ±μdg are the limits of the power cal vehicles. As is shown in (6), the micro-grid system has
increment. The dynamic of the DG system can be expressed been described as the state-space model. And it can be trans-
as: formed into the following discrete state space model:
1 1
ΔṖDG = ΔXG − ΔPDG (3)
Td Td x(k + 1) = Ax(k) + Bu u(k) + Bd d(k) (7a)
yc (k) = Cc x(k) (7b)
1 1 1
ΔẊG = ΔuDG − ΔXG − Δf (4)
Tg Tg RTg To reduce or eliminate static errors, formula (7) can be
In the actual system, the output power increment and its transformed into the following form:
change rate of DG should satisfy the following constraints:
Δx(k + 1) = AΔx(k) + Bu Δu(k) + Bd Δd(k) (8a)
− μdg  ΔPDG (t)  μdg (5a) yc (k) = Cc Δx(k) + yc (k − 1) (8b)
− δdg  ΔṖDG (t)  δdg (5b)
where,

In summary, the state space of the isolated micro-grid with Δx(k) = x(k) − x(k − 1) (9a)
electrical vehicles can be denoted as follows:
Δu(k) = u(k) − u(k − 1) (9b)

ẋ(t) = Ac x(t) + Bcu u(t) + Bcd d(t) (6a) Δd(k) = d(k) − d(k − 1) (9c)

y = Cc x(t) (6b) Assuming that p is the finite prediction horizon, m is the

control horizon and m  p.
and
At time k, choosing Δx(k) = x(k) − x(k − 1) to be the
⎡ 1 1 1
⎤ new state input variable. From ( 12), the future state is shown
0 2H t 0 2H t 2H t
⎢ ⎥ as follow:
⎢ 0 − T1d 1
0 0 ⎥
⎢ Td ⎥
Ac = ⎢ 1
⎢− RT g 0 − T1g 0 0 ⎥
⎥ Δx(k + 1|k) = AΔx(k) + Bu Δu(k) + Bd Δd(k)
⎢ ⎥
⎣ 0 0 0 − T1e1 0 ⎦ Δx(k + 2|k) = AΔx(k + 1|k) + Bu Δu(k + 2)
0 0 0 0 − T1e2
⎡ ⎤ + Bd Δd(k + 2)
0 0 0 ..
⎢ ⎥
⎢0 0 0 ⎥ .
⎢1 ⎥
⎢
Bcu = ⎢ Tg 0 0 ⎥ Δx(k + m|k) = AΔx(k + m − 1|k) + Bu Δu(k + m − 1)
⎥
⎢ ⎥
⎣ 0 T1e1 0 ⎦ + Bd Δd(k + m − 1)
1
0 0 Te2 ..
 T .
1
Bcd = − 2H 0 0 0 0

t
Δx(k + p|k) = AΔx(k + p − 1|k) + Bu Δu(k + p − 1)
Cc = 1 0 0 0 0 + Bd Δd(k + p − 1)

3590
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.
 Similarly, the controlled output of k + 1 to k + p is shown where,
as below:

p
2
yc (k + 1|k) = Cc Δx(k + 1|k) + yc (k) J(x(k), ΔU (k), m, p) = (yc (k + i|k) − r(k + i|k))Q
i=1
yc (k + 2|k) = Cc Δx(k + 2|k) + yc (k + 1|k)

m
2
.. + Δu(k + i|k)R
. i=1
(14)
yc (k + m|k) = Cc Δx(k + m|k) + yc (k + m − 1|k)
In (14), yc (k + i|k) is the predicted controlled output;
..
. r(k + i|k) is the setpoint; Δu(k + i|k) is the predicted
yc (k + p|k) = Cc Δx(k + p|k) + yc (k + p − 1|k) control increment; p is a finite prediction horizon, m is the
control horizon, and for simplicity of exposition, the con-
Transform the above results into a matrix form, defining
trol and predictive horizons are chosen to have identical val-
the p step prediction output and m step input as follow:
⎡ ⎤ ues. Q ∈ Rn×n and R ∈ Rm×m denote positive definite
yc (k + 1|k) and symmetric weighting matrices. They are tuning param-
⎢ ⎥
⎢yc (k + 2|k)⎥ eters to achieve the desired performance and can be cho-
YP (k + 1|k) = ⎢ ⎢ ⎥ (10)
... ⎥
⎣ ⎦ sen freely. The weighting matrices Q and R in the objec-
yc (k + p|k) p×1 tive function (10) are chosen as Q = diag(1, 0, 0, 0, 0) and
⎡ ⎤ R = diag(1, 1, 1).
Δu(k)
⎢ ⎥ In general, in order to avoid the system parameters change
⎢ Δu(k + 1) ⎥
ΔU (k) = ⎢⎢ ⎥ (11) too fast causing irreversible damage to the equipments, the
.. ⎥
⎣ . ⎦ limitation of the output power increment and the output pow-
Δu(k + n − 1) m×1 er increment rate of each subsystem should be considered.
The output of the future p step prediction can be derived The constraints of equivalent EVs and DG system can be
from the following prediction equation: formulated as follows [16]:

YP (k + 1|k) = ΦΔx(k) + Γyc (k) + EΔd(k) + HΔU (k) − μdg  ΔPDG (t)  μdg (15a)
(12)
− δdg Ts  ΔPDG (t) − ΔPDG (t − 1)  δdg Ts (15b)
where,
⎡ ⎤ − μej  ΔPEj (t)  μej , j = 1, 2 (15c)
Cc A ⎡ ⎤
⎢
2 ⎥ Inc ×nc − δej Ts  ΔPEj (t) − ΔPEj (t − 1)  δej Ts , j = 1, 2
⎢ Cc Ai ⎥ ⎢ ⎥
⎢ ⎥ ⎢Inc ×nc ⎥ (15d)
⎢i=1 ⎥ ⎢ . ⎥
Φ=⎢ .. ⎥ Γ = ⎢ . ⎥
⎢ ⎥ ⎣ . ⎦
⎢ . ⎥ where Ts is the sampling interval. These constraints can be
⎣
p ⎦
C Ai Inc ×nc p×1 expressed as the following form:
c
i=1 p×1
⎡ ⎤ LΔU ≤ V (16)
C c Bd
⎢
2 ⎥
⎢ Cc Ai−1 Bd ⎥
⎢ ⎥ To get optimal control signals, we can solve a standard QP
⎢i=1 ⎥
E=⎢ .. ⎥ problem with constraints as follows:
⎢ ⎥
⎢ . ⎥
⎣
p ⎦
Cc Ai−1 Bd min J(x(k), ΔU (k), m, p) (17a)
ΔU (k)
i=1 p×1
⎡ C c Bu 0 ··· 0 ⎤ s.t. LΔU ≤ V (17b)

2
⎢ Cc A Bu i−1
0 ··· 0 ⎥
⎢ i=1 ⎥
⎢ .. .. .. .. ⎥ 4 Simulation and analysis
⎢ . ⎥
⎢ m . . . ⎥
H=⎢ 
⎢ Cc Ai−1 Bu

m−1
Cc Ai−1 Bu ··· C c Bu
⎥
⎥ In this section, we firstly investigate the system response
⎢ i=1 i=1 ⎥
⎢ .. .. .. ⎥ to step load changes. The simulations for MPC is performed
⎢ .. ⎥
⎣ . . . . ⎦ with a sampling time of Ts = 1s and prediction horizon of

p 
p−1 
p−m+1
Cc Ai−1 Bu Cc Ai−1 Bu ··· Cc Ai−1 Bu
i=1 i=1 i=1 p×m
p = 15s. The parameters of micro-grids are listed in Ta-
And the MPC for the overall system (7) solves the follow- ble 1 [19].
ing optimization problem: In this scenario, performance of the proposed MPC with
respect to step load disturbance is investigated. In order to
min J(x(k), ΔU (k), m, p) (13) simulate disturbance on system, 5% load increase is applied
ΔU (k)

3591
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.
 −3 −3
x 10 x 10
5 20

ΔPE2
10
Δf

−5
0
−10
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
(a) (f)
0.06 0.02

ΔṖE2
ΔPDG

0.04 0.01
0.02
0
0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
(b) (g)
−3
x 10
4 0.08
0.06
ΔṖDG

ΔuDG
2
0.04
0 0.02
0
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
(c) (h)
0.04 0.03
0.02
ΔPE1

ΔuE1
0.02
0.01
0 0
−0.01
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
(d) (i)
0.03
0.02 0.02
ΔṖE1

ΔuE2
0.01 0.01

0 0
−0.01
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
(e) (j)
Time (sec) Time (sec)

Fig. 4: Response to step load disturbance of the micro-grid with EVs.

Table 1: Parameters of micro-grid system for 5% load increase in DG system. What’s more, both the
Micro-grids Component Parameters Values Unit output increment and the output power increment rate are
Tg 0.1 s satisfy the constraints. It’s obvious that the proposed MPC
Td 8 s controller can effectively solve the LFC problem of micro-
DG R 2.5 Hz/pu MW grid with electrical vehicles.
δdg 0.0018 pu MW/s
μdg 0.06 pu MW 5 Conclusions
Te1 1 s
In this paper we have formulated the LFC problem of
δe1 0.01 pu MW/s
micro-grid system with electrical vehicles as a disturbance
EV1 μe1 0.025 pu MW
rejection problem with constraints, that represent require-
Emax 0.95 pu MWh
Emin 0.80 pu MWh
ments on output power increment and the output power in-
Te2 1 s crement rate within bounds. The MPC is applied to the LFC
δe2 0.01 pu MW/s problem of the micro-grid system with step loads. Analy-
EV2 μe2 0.015 pu MW sis and simulation results have confirmed the benefits of the
Emax 0.90 pu MWh designed MPC controller in driving the frequency deviation-
Emin 0.80 pu MWh s Δf to zero, while respecting the equivalent EVs and DG
system constraints.

References
in DG system. The load disturbance rejection performance
of the MPC formulation is evaluated. The relative perfor- [1] D. Tran, Energy Management for Lifetime Extension of Ener-
mance of MPC rejecting the load disturbance in DG system gy Storage System in Micro-Grid Applications, IEEE Transac-
tions on Smart Grid, 4(3): 1289–1296, 2013.
is depicted in Fig. 4. Fig. 4(a) shows the frequency devia-
[2] M. S. Mahmoud, N. M. Alyazidi, and M. I. Abouheaf, Adap-
tion of the micro-grid with EVs. Fig. 4(b) and (c) show the
tive intelligent techniques for microgrid control systems: A
output power increment and the output power increment rate
survey, International Journal of Electrical Power & Energy
of the DG system. Fig. 4(d), (e) and (f), (g) show the out- Systems, 90: 292–305, 2017.
put power increment and the output power increment rate of [3] H. Y. Guo, F. Liu, R. Yu, R. Yu, Z. P. Sun, and H. Chen,
the EV1 and EV2, respectively. Fig. 4(h), (i) and (j) show Regional path moving horizon tracking controller design for
the control input of the microgrid with EVs. With the M- autonomous ground vehicles, Science China(Information Sci-
PC framework, frequency deviation is rejected in about 60s ences), 60(1): 013201, 2017.

3592
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.
 [4] M. S. Elnozahy, and M. M. A. Salama, Studying the feasibility Springer Science Business Media, 2013.
of charging plug-in hybrid electrical vehicles using photovolta- [19] J. Yang, L. He, and S. Fu, An improved PSO-based charg-
ic electricity in residential distribution systems, Electric Power ing strategy of electric vehicles in electrical distribution grid,
Systems Research, 110(110): 133–143, 2014. Applied Energy, 128(3): 82–92, 2014.
[5] L. E. Bremermann, M. Matos, J. A. P. Lopes and M. Rosa,
Electric vehicle models for evaluating the security of supply,
Electric Power Systems Research, 111(6): 32–39, 2014.
[6] M. Singh, P. Kumar, and I. Kar, Implementation of Vehicle to
Grid Infrastructure Using Fuzzy Logic Controller, IEEE Trans-
actions on Smart Grid, 3(1): 565–577, 2012.
[7] M. H. Khooban, T. Niknam, F. Blaabjerg and T. Dragicevic, A
new load frequency control strategy for micro-grids with con-
sidering electrical vehicles, Electric Power Systems Research,
143: 585–598, 2017.
[8] S. Pati,K. B. Mohanty, S. K. Kar, and S. Mishra, Voltage and
frequency control of a micro-grid using a fuzzy logic controller
based STATCOM equiped with battery energy storage system,
in International Conference on Circuit, Power and Computing
Technologies, pp. 1–7, 2016.
[9] S. Leonori, E. D. Santis, A. Rizzi and F. M. F. Mascioli, Multi
objective optimization of a fuzzy logic controller for energy
management in microgrids, in Evolutionary Computation, pp.
319–326, 2016.
[10] S. Pati, S. K. Kar, K. B. Mohanty, and S. Mishra, A sliding
mode controller based STATCOM with battery storage for volt-
age and frequency stabilization in a micro-grid, in Internation-
al Conference on Circuit, Power and Computing Technologies,
pp. 1–7, 2016.
[11] C. Mu, Y. Tang, and H. He, Observer-based sliding mode fre-
quency control for micro-grid with photovoltaic energy inte-
gration, in Power and Energy Society General Meeting, 2016.
[12] H. T. Zhang, Z. H. Chen, Y. J. Wang, M. Li, and T. Qin, Adap-
tive predictive control algorithm based on Laguerre Function-
al Model, International Journal of Adaptive Control & Signal
Processing, 20(2): 53-76, 2006.
[13] M. H. Khooban, T. Niknam, F. Blaabjerg, P. Davari and
T. Dragicevic, A robust adaptive load frequency control for
micro-grids, Isa Transactions, 65: 220, 2016.
[14] X. J. Liu, L. X. Niu, and J. Z. Liu, Nonlinear multivariable
supervisory predictive control, pp. 2779–2784, 2009.
[15] P. S. V. Sagar, and K. S. Swarup, Load frequency con-
trol in isolated micro-grids using centralized model predic-
tive control, in Power Electronics, Drives and Energy Systems
(PEDES), 2016 IEEE International Conference on, pp. 3–16,
2016.
[16] J. Yang, Z. Zeng,Y. Tang,J. Yan,H. He, and Y. Wu, Load fre-
quency control in isolated micro-grids with electrical vehicles
based on multivariable generalized predictive theory, Energies,
8(3): 2145–2164, 2015.
[17] Z. M. Cheng, H. T. Zhang, M. C. Fan, and G. R. Chen, Dis-
tributed Consensus of Multi-Agent Systems With Input Con-
straints: A Model Predictive Control Approach, IEEE Transac-
tions on Circuits & Systems I Regular Papers, 62(3): 825–834,
2015.
[18] E. F. Camacho, and C. B. Alba, Model predictive control,

3593
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 10,2020 at 06:46:51 UTC from IEEE Xplore. Restrictions apply.