Send Orders for Reprints to reprints@benthamscience.

ae

244 The Open Fuels & Energy Science Journal, 2015, 8, 244-251

Open Access
The PV MPPT Control Strategy Based on Generalized Predictive Control
Ni Fuyin1,2, Li Zhengming*,1 and Shen Lin2

1
School of Electrical and Information Engineering, Jiangsu University, Jiangsu Zhengjiang, 212013, P.R. China
2
School of Electrical and Information Engineering, Jiangsu University of Technology, Jiangsu Changzhou, 213001, P.R.
China

E
Abstract: The excursion of maximum power point always exists due to nonlinear and time-varying characteristics of

L
photovoltaic (PV) generation. To solve this problem, the maximum power point tracking (MPPT) control strategy based
on generalized predictive control (GPC) is put forward. Firstly, the theory and feature of PV cell is analyzed, and the

C
MPPT control model is constituted based on Boost circuit. Then the MPPT algorithm based on GPC is illustrated from

I
such sections as prediction model, rolling optimization, online identification and feedback correction, the algorithm steps
are given as well. Besides, the MPPT simulations are conducted by the methods based on GPC and PID in comparison.

T
The comparison results testify that GPC algorithm is faster and more accurate to track the maximum power point. GPC
algorithm has the characteristics like stability and robustness, and can be applied to solar generation system.

R
Keywords: Boost circuit, generalized predictive control, MPPT, photovoltaic cell, PID.

A
1. INTRODUCTION operation. The PID method in P&O is to feedback and
correct deviation depending on of history error data. So the
With the industrialized society transforming to low- tracking precision and speed of response has contradiction to

D
carbon society in the current world, developing green energy a certain degree, and the system cannot achieve the optimal
has become the research focus in all countries. Solar power control result.

E
generation has cleanable and renewable characteristics,
which is recognized publicly as the most potential new In order to overcome the insufficient of PID method,

T
energy technique. The source of solar generation is PV array, MPPT improved method based on generalized predictive
and the output power has a maximum power point in control (GPC) is proposed in the article. Firstly, PV cell
theory and the equivalent circuit based on Boost circuit are

C
different external environment like solar irradiation,
temperature, and variable load etc. But the nonlinearity and to be analyzed, the maximum power express is to be deduced
time-varying characteristics result in the excursion of as well. Then the MPPT control system based on GPC is to

A
maximum power point. The research centralized on be presented from the processes including prediction model,
maximum power point tracking (MPPT) is conducted rolling optimization, and online feedback correction. Next,

R
extensively, and many achievements are obtained [1-12]. the optimum control algorithm is to be realized according to
PV array dynamic feature. Eventually, the simulation

T
Most of MPPT control strategies are combinations comparison analysis is to be performed by Simulink between
algorithms considering some intelligent control algorithm, or PID method and GPC method.

E
some improved strategies based on the classical algorithm.
Generally, the researches on PV MPPT control strategies are
mainly classified as below: (1) Constant voltage method. 2. ANALYSIS OF PV CELL THEORY

R
This method is simple and effective, but the accuracy is not The math model of PV cell is presented in Eq. (1) [12]
very high, and the power loss is relative much. (2)
Conductance increment method [3, 4]. This method avoids ⎧ q(U + IRs ) ⎫ U + IRs
misjudgment like constant voltage method, and can output I = I ph − I o ⎨exp[ − 1⎬ − (1)
⎩ AKT ⎭ Rsh
stable voltage when tracking solar irradiation. Its drawback
is that sensor needs high precision and system response where,
needs high speed. Therefore the hardware cost of inverter is I Photovoltaic current (A)
higher than the other methods. (3) Perturbation and
observation (P&O) method [5-11], this method is the most I ph Photo current (A)
applied and studied for its good performance and simple Io Diode saturation current (A)

A Idealist factor for a p-n junction

Boltzman's factor (1.381 × 10 J/K )
-23
K

1876-973X/15 2015 Bentham Open

The PV MPPT Control Strategy Based on Generalized Predictive Control The Open Fuels & Energy Science Journal, 2015, Volume 8 245

T Temperature of a solar array ( oC) R pv IL

Rs Series resistance (Ω) + +
BOOST
Rsh Shunt resistance (Ω) U pv MPPT RL UL
_ _
-19
q Charge of an electron (1.602 × 10 C)
U Photovoltaic voltage (V) Rpv
IO
+
The engineering math model of PV cell in the practical +
application is expressed in Eq. (2) U pv Req UO
_ _
U
I = I sc {1 − C1[exp( ) − 1]} (2)

E
C2U oc
Fig. (2). Equivalent transformation of boost circuit.
−U m where, R is the resistance of PV array. As shown in Fig. (2),

L
Im
where, C1 = (1 − ) exp( ),
pv

I sc C2U oc the equivalent resistance is deduced in Eq. (5) in accord with

C
Thevenin’s theorem:

I
Um Im −1
C2 = ( − 1)[ln(1 − )] Req = RL ⋅ (1- D) 2
(5)
U oc I sc

T
Therefore the value of input resistant is adjusted by
I sc Short-circuit current (A) means of regulating duty cycle of Boost circuit. When

R
U oc Open-circuit voltage (V) equivalent input impedance Req is matched Rpv by adjusting
Im Current at the maximum power point (A) duty cycle, the PV array outputs maximum power expressed

A
Um Voltage at the maximum power point (V) in Eq. (6):
U PV U 2
And the four technical parameters are provided in the Pmax = ( )2 ⋅ Req = PV (6)

D
standard condition by PV manufacturer. RPV + Req 4Req

E
3. ANALYSIS OF THE BOOST EQUIVALENT 4. MPPT CONTROL SCHEME BASED ON GPC
CIRCUIT

T
GPC has the characteristics such as predictive model,
The solar inverter uses two-stage structure. The front- rolling optimization and feedback correction on line etc [13,

C
stage is Boost circuit, which realizes the functions of 14]. So GPC fuses the advantages of predictive control and
voltage-rise and MPPT. The back-stage is grid-connected self-adaptive control, and easily estimates online parameters

A
inverter, which achieves the function of inverting-inversion and realizes self-adaptive control. By means of rolling
and grid connection. The Boost schematic diagram is as optimization, identification and feedback correction, GPC
below in Fig. (1).

R
can estimate online model parameters according to the actual
continuous input and output information. What’s more, it can
IL IO

T
correct the control law to improve the control performance in
different external environment.
+ L +

E
VD
U pv VT UO RL
The PV MPPT scheme based on GPC in Boost model is
shown in Fig. (3). In the MPPT module, the P&O method is
_ _

R
adapt to output reference voltage U by the detection of PV
ref

Fig. (1). Boost schematic diagram. array voltage and current output, and the U subtracts U to ref pv

According Fig. (1), the relation between output voltage get the dynamic voltage error U . GPC replaces traditional
err

U pv and U is: o
PID and regulate the error to output switch signal which is to
control PWM module. Eventually, the output power of
U pv ⋅ D ⋅ T + (U pv -U o ) ⋅ (1 -D ⋅ T )=0 (3) photovoltaic array is kept maximum according to the
switching tube status controlled by PWM.
where, T is switching period of switching tube, D is duty
cycle, and its interval is (0,1). 5. GPC ALGORITHM IN MPPT STRATEGY
According to Eq. (3), the output voltage U can be o
5.1. GPC Structure
expressed in Eq. (4):
GPC algorithm structure is shown in Fig. (4) [15]. It takes
1
Uo = U pv (4) the controlled auto-regressive moving average (CARIMA) as
1-D the predictive control model and describes the dynamic relation.
Equivalent transformation is employed in the Boost Also, the identified model parameter and information are
circuit, Fig. (1) is depicted as Fig. (2): provided to the GPC optimization algorithm in real time.
246 The Open Fuels & Energy Science Journal, 2015, Volume 8 Fuyin et al.

L E
I C
T
Fig. (3). PV MPPT control based on GPC in Boost modeling.

R
random GPC control block diagram is shown in Fig. (5) [16]. It is
noise

A
composed of the process including softening, adjustment,
ξ(k) and prediction. Note that the expected output vector is
yr (k ) GPC u (k ) y (k ) obtained after set-sequence is softened at every moment.
optimal Then the deviation vector is got after the expected output

D
Online
algorithm
identficiation vector compared to the predictive output. The control
CARIMA

E
increment Δu(k) is got by dot product between deviation
model
Predictive
ym ( k ) vector and dynamic vector g at the moment. On one hand,
T

T
output
the control increment acts on control object by the
Fig. (4). GPC algorithm structure. calculation of digital integration. On the other hand, it

C
predicts new system output f with the system output.
Then the reference voltage is calculated by the algorithm.
u(k) is input signal and equal to U ref − U pv in the Boost circuit.

A
5.2. Prediction Model
y(k) is output signal, and equal to Vswt which control PWM,

R
yr (k) is set-error of U ref − U pv . CARIMA applied in the minimum variance control is
used to describe basic prediction model of system dynamic

E T softening control preidiction control

ξ (k )
C

R
M object A⋅ Δ
adjustment
⎛1 − a ⎞ W gT
yr(k) ⎜ ⎟ + + Δu (k ) u (k ) + + y (k )
⎜1 − a2 ⎟ 1 z −1 B
⎜ ⎟ [ d1d 2  d n ] 1 − z −1
 A
⎜⎜ ⎟ +
⎝1 − a n ⎟⎠ −

⎛ g z −1+ ⎞
⎛ a1 ⎞ ⎜ 11 ⎟
⎜ ⎟ H ⎜  ⎟
Q ⎜  ⎟ ⎜ ⎟
⎜ an ⎟ ⎜ g z −1+ ⎟
⎝ ⎠ ⎝ nn ⎠ F

+ ⎛ F1 ⎞
f + ⎜ ⎟
⎜  ⎟
⎜ F ⎟
⎝ n ⎠

Fig. (5). GPC algorithm control block diagram.

The PV MPPT Control Strategy Based on Generalized Predictive Control The Open Fuels & Energy Science Journal, 2015, Volume 8 247

action. It has prediction function, and can predict future P M

output according the history data and future input. The future min J (k) = E{ ∑ [y m
(k + 1) − yr (k + 1)] +
2
∑ λ [Δu(k + j − 1)] }
j
2
(12)
j =1 j =1
system output is shown in Eq. (7):
−1
where, P is prediction time domain, M is control time
C(q )ξ (k)
−1 −1
A(q )y(k) = B(q )u(k − 1) + (7) domain, λ j is control increment weighted parameter, which
Δ

where, A(q −1 ) = 1 + a1q −1 +  + an q − n ,

is set to constant commonly. y (k + j) is reference trajectory.
r

−1
B(q ) = b0 + b1q +  + bm q −1 −m
. Therefore, the expected output takes the form of reference
trajectory so that the controlled object output y(k) follows a
where, q −1 is backward operator, −1
y(k)q = y(k − 1) , specified soften curve. Eventually, yr (k) reaches the target
−1
Δ = 1− q is difference operator. ξ (k) is the single random

E
W smoothly.
noise sequence that mean value is 0. According to previous Define:

L
analysis, u(k) and y(k) is input and output of controlled
E j BΔu(k + j − 1) = G j1Δu1 + G j 2 Δu2 (13)
object respectively. Note that C(q −1
)=1 , and n is order of

C
input, and m is order of output. where, G j1 = [g j , j−1  g j ,0 ] , G j 2 = [g j , j+m−1 g j , j ],

I
The system model at the k + j moment is
Δu1 = [Δu(k)  Δu(k + j − 1)]T ,

T
ξ (k + j ) Δu2 = [Δu(k − m)  Δu(k − 1)]T .
−1 −1
A(q )y(k + j ) = B(q )u(k + j − 1) + (8)
Δ According to Eq. (11) and (13), Eq. (14) is obtained:

R
For y(k + j) exists unknown information, the ym (k + j) = G j1Δu1G j 2 Δu2 + Fj y(k) (14)

A
Diophantine equation is introduced and the output predicted
value is obtained at the k + j moment. The Diophantine ⎡ g1 0 ⎤
⎢ ⎥
equation is presented in Eq. (9) ⎡G11 ⎤ ⎢ g2 g1 ⎥

D
−1 −1 −j −1 ⎢G ⎥ ⎢ ⎥
1 = A(q )E j (q )Δ + q F j (q ) (9) 
Define: G = ⎢
21 ⎥
=⎢ ⎥,
⎢  ⎥ ⎢ gM −1 gM −1  ⎥

E
g1
where, E j (q −1 ) = e j,0 + e j,1q −1 ++ e j, j−1q −( j−1) , ⎢ ⎥ ⎢ ⎥
⎣GP1 ⎦ ⎢     ⎥

T
Fj (q −1 ) = f j,0 + f j,1q −1 ++ f j,n q − n . ⎢ gp g p−1  gP−M +1 ⎥
⎣ ⎦

C
−1
E j and Fj is determined by both A(q ) and prediction ⎡G12 Δu2 + F1 y(k) ⎤
length j . Then the simplified equation is got in Eq. (10): ⎢G Δu + F y(k) ⎥
f =⎢ ⎥

A
22 2 2

y(k + j) = BE j Δu(k + j − 1) + Fj y(k) + E jξ (k + j) (10) ⎢  ⎥

⎢ ⎥
⎣GP 2 Δu2 + FP y(k) ⎦

R
Note that the future noise is unknown, but its mean is 0.
So GPC prediction model at the k + j moment is Considering control time domain defined as M ,

T
ym (k + j) = G j Δu(k + j − 1) + Fj y(k) (11) Define:

E
U = Δu1 = [Δ(k)  Δ(k + j − 1)] = [Δ(k)  Δ(k + M − 1)]
T T

−j
B(1 − q Fj )
where, G = BE = −1
= g j ,0 + g j ,1q +  + g j ,m+ j −1q
− ( m+ j −1)
.

R
j j Ym = [ym (k + 1) ym (k + 2)  ym (k + P)]
AΔ
where, U is future controlled variable, Ym is future output.
5.3. Rolling Optimization The model of prediction control is changed to vector form:
The optimization of prediction control is a kind of limited Ym = GU + f (15)
time domain rolling optimization, i.e., the optimized
performance indicator involves the moment and a future Also, Eq. (10) is changed to vector with Eq. (11)
limited time. Till the next moment the optimization period substituted into the equation, then Eq. (16) is got:
moves head. So the optimization is not conducted once time
off line, but conducted continuously on line, that is so called J = (GU + f − Yr ) (GU + f − Yr ) + U U
T T
(16)
rolling optimization. Calculate the derivative by the future variable U , set the
GPC uses quadric performance indicator that the output results equals to 0, the following express is obtained:
error and the control increment are weighted: −1
U = (G G + λ I ) G (Yr − f )
T T
(17)
248 The Open Fuels & Energy Science Journal, 2015, Volume 8 Fuyin et al.

where, Y = [y (k + 1) y (k + 2)  y (k + P)]
r r r r
(4) Construct vector matrix G and f .
The optimal controlled instant variable is (5) Calculate and perform control u(k) according to Eq.
(18).
Δu(k) = [1 0  0](G G + λ I ) G (Yr − f )
T −1 T
(18)
(6) Continue to circulation back to step (3).
Then, u(k) = u(k − 1) + Δu(k) .
6. EXPERIMENT AND ANALYSIS
5.4. Online Identification and Feedback Correction 6.1. PV Array MPPT Simulation and Analysis at
The future control effect is determined by optimization. Different Time
The prediction control only takes effect at this moment, till
The photovoltaic Boost modeling is depicted in Fig. (6),
the next moment, the control system checks the output of the MPPT algorithm based on GPC is simulated by
object, and corrects the prediction model control by the

E
Simulink. The simulation parameters are as below:
previous output, then begins the new optimization.
U oc = 22V , I sc = 8.58A , U m = 17.7V , I m = 7.94A , Pmax is about 140

L
GPC control algorithm uses one control model, and keeps
the prediction correctness by online feedback. W.

C
Δy(k) in Eq. (9) is changed to Eq. (19): Comprehensive considering MPPT control effect after

I
simulating some times, the parameters in GPC algorithm are
Δy(k) = (1 − A)y(k) + BΔu(k − 1) + ξ (k) (19) given as below:

T
And model parameter (A, B) in Eq. (19) and data Calculation time period is 500 µs , control time domain
parameter (y,u) is defined as vector θ and ϕ k respectively. P = 8 , prediction time domain M = 4 , weighted parameter

R
k

λ = 1.2 , input order n = 2 , output order m = 3 , forgetting

where, θ k = [a1 an | b1 bn ]
factor β = 0.99 .

A
ϕ k = [Δy(k − 1) − Δy(k − n) | −Δu(k − 1) − Δu(k − m)]
T

Suppose the solar irradiation is 600W/m , temperature is

2

Therefore Eq. (16) is changed to : 20℃. Note that the methods of MPPT are compared between

D
Δy(k) = ϕ k θ k + ε (k)
T
(20) PID and GPC, the Pout output waves are shown in Fig. (6a,
b) according to different control methods:

E
The recursive least-squares (RLS) method with forgetting
(a) by the method of PID
factor is used to estimate model parameter θ k :
Pout

T
θ (k) = θ (k − 1) + K (k)[y(k) − ϕ (k)θ (k − 1)]
T
(21)

C
K (k) = P(k − 1)ϕ (k)[ϕ (k)P(k)ϕ (k) + µ ]
T −1
(22)

A
1
P(k ) = [ I − K (k )ϕ (k )]P(k − 1)
T
(23)
µ

R
where, µ is forgetting factor, usually, 0.95 < µ ≤ 1, K (k) is

T
weight factor, P(k) is positive definite covariance matrix.
The initial value of parameter vector and covariance matrix

E
should be set when to begin control, the common setting
method is depicted in Eq. (24):

R
⎧ P(0) = α I , α is sufficently larege real number
2

⎨ (24)
⎩θ (0) = ε , ε is sufficinently small real vector (b) Pout by the method of GPC
In conclusion, the online operation step of GPC
algorithm is as below:
(1) The parameter !n, m, P, M , λ , µ !and!α is given in
according with controlled object and control request of the
Boost circuit.
(2) Set P(0) and θ (0) initial value, input preliminary data.
(3) Use RLS to estimate the model parameter !A(q ) and −1

) , solve recursively Diophantine equation and get Fj and

−1

!B(q
Gj .

Fig. (6). MPPT control simulation when irradiation is 600W/m

2
.
The PV MPPT Control Strategy Based on Generalized Predictive Control The Open Fuels & Energy Science Journal, 2015, Volume 8 249

(a) Pout by the method of PID

(b) Pout by the method of GPC

L E
I C
R T
A
E D
T
Fig. (7). MPPT control simulation when irradiation is 1000W/m
2
.
(a) Pout by the method of PID

A C
T R
RE (b) Pout by the method of GPC

Fig. (8). MPPT simulation in the abrupt change of solar irradiation.

250 The Open Fuels & Energy Science Journal, 2015, Volume 8 Fuyin et al.

Suppose the solar irradiation is 1000W/m 2 , temperature is ACKNOWLEDGEMENTS

25℃, the simulation figures are shown in Fig. (8a, b) based Supported by National Natural Science Foundation of
on the method of PID and GPC: China (51477070).
From Fig. (7), it is concluded that the method on GPC is Supported by Jiangsu Province Prospective Joint
more accurate and faster to track the maximum power point Research Project (BY2015028-01).
than the method on PID. So GPC method decreases the loss
and improves control stability and rapidity compared to PID Supported by Changzhou Key Laboratory of The Internet
method. of Things New Technology and Application (CM20133008).
Supported by Natural Science Foundation of Jiangsu
6.2. MPPT Simulation in the Abrupt Change of Solar University of Technology (KYY14001).
Irradiation Supported by A Project Funded by the Priority Academic

E
The simulation is carried out to test the adaptive ability Program Development of Jiangsu Higher Education
Institutions (PAPD).

L
and dynamic response ability of MPPT algorithm by sudden
change of solar irradiation. Note that the solar irradiation is
REFERENCES

C
separately set to 1000 W/m 2 in 0~0.2 s, 600 W/m 2 in

I
0.2~0.4 s, and 800 W/m 2 in 0.4~0.6 s. The MPPT simulation [1] Sundareswaran, K.; Peddapati, S.; Palani, S. MPPT of PV systems
figures are shown in Fig. (8a, b) by the methods of PID and under partial shaded conditions through a colony of flashing

T
fireflies. IEEE Trans. Energy Conver., 2014, 29, 463-472.
GPC. [2] Mamarelis, E.; Petrone, G.; Spagnuolo, G. Design of a sliding-
It is observed that the solar irradiation occurs sudden mode controlled SEPIC for PV MPPT applications. IEEE

R
Transactions on Industr. Electron., 2014, 61, 3387-3398.
change at 0.2 s and 0.4 s from the dynamic response curve in [3] Hsieh, G.C.; Tsai, C.Y.; Wang, C.H. Photovoltaic power-increment
Fig. (8). The method based on GPC can adjust maximum B aided incremental-conductance MPPT with two-phased tracking.

A
power point rapidly, and keep PV array output maximum IEEE Trans. Power Electro., 2013, 28, 2895-2911.
power. GPC has better dynamic and stability feature, and [4] Hsieh, G.C.; Hsieh, H.I.; Tsai, C.Y.; Wang, C.H. Photovoltaic
power increment-aided incremental-conductance MPPT with two-
improves control performance facing to various external
phased tracking. IEEE Trans. Power Electron., 2013, 28, 2895–

D
environment at the comparison of PID. What’s more, the 2911.
GPC algorithm can keep high efficiency to track maximum [5] Jiang, Y.C.; Qahouq, J.A.A.; Haskew, T.A. Adaptive step size with

E
power point and operate in the optimal status. adaptive-perturbation-frequency digital MPPT controller for a
single-sensor. IEEE Trans. Power Electron., 2014, 28, 3195-3205.
[6] Lee, K.J.; Kim, R.Y. An adaptive maximum power point tracking

T
CONCLUSION scheme based on a variable scaling factor for photovoltaic. IEEE
Trans. Energy Conver., 2012, 27, 1002-1008.
Aiming at the existing deviation of PV array maximum

C
[7] Chen, Y.A.; Zhou, J.H.; Li, J.; Li, L.L. Application of gradient
power point in different external environment, MPPT variable step size MPPT algorithm in photovoltaic system. Proc.
improved strategy based on GPC is proposed. The model of CSEE, 2014, 34, 3156-3161.

A
[8] Kollimalla, S.K.; Mishra, M.K. A novel adaptive P&O MPPT
PV MPPT control based on GPC in Boost is made. Then the algorithm considering sudden changes in the irradiance. IEEE
control express is deduced, the algorithm step is given as Trans. Energy Conver., 2014, 29, 602-610.

R
well. Next, the simulations between methods of GPC and [9] Liu, S.J.; Zhuang, S.X.; Chen, X. Research on the control method
PID control are given in different time and in the state of of current disturbance observers for photovoltaic grid-connected

T
inverters. Proc. CSEE, 2014, 33, 5811-5818.
abrupt solar change respectively. The results indicate that [10] Sera, D.; Mathe, L.; Kerekes, T.; Spataru, S.V.; Teodorescu, R. On
GPC method shortens transient time clearly and increases the perturb-and-observe and incremental conductance MPPT

E
disturbance-resisting ability, so the method improves the methods for PV systems. IEEE J. Photovoltaics, 2013, 3, 1070-
solar generation efficiency at the comparison of PID. Hence, 1078.
[11] de Brito, M.A.G.; Galotto, L.; Sampaio, L.P.; de Azevedo e Melo,

R
the MPPT algorithm based on GPC provides a new strategy
G.; Canesin, C.A. Evaluation of the main MPPT techniques for
to solar generation research. photovoltaic applications. IEEE Trans. Industr. Electron., 2013, 60,
1156-1167.
[12] Wang, H.B.; He, J.H. Application of conjugate gradient method in
CONFLICT OF INTEREST
the photovoltaic system. J. Tsinghua Univ. (Sci. Technol.), 2014,
53, 269-273.
The authors confirm that this article content has no [13] Xu, F.; Wei, X.L.; Ren, L.H.; Luo, X.L. A control structure
conflict of interest. selection method based on multivariable generalized predictive
control for unstable processes. Acta Automat. Sin., 2013, 39, 1547-
1551.
The PV MPPT Control Strategy Based on Generalized Predictive Control The Open Fuels & Energy Science Journal, 2015, Volume 8 251

[14] Shi, W.G.; Shao, C.; Sun, Z.Y. Improved GPC network-control [16] Sun, J.; Zhang, H.Y.; Li, X.; Zhang D.H. Generalized predictive
algorithm based on AR model time-delay prediction. Contr. control scheme for monitor AGC system. J. Central South Univ.
Decision, 2012, 27, 477-480. (Sci. Technol.), 2012, 43, 3852-3856.
[15 ] Liu, Y.M.; Hua, Z.Y.; Chen, Y.Q. The application of GPC
algorithm to the air-fuel ratio control of engine. Automot. Eng.,
2013, 35, 403-407.

Received: April 17, 2015 Revised: May 21, 2015 Accepted: May 26, 2015

© Fuyin et al.; Licensee Bentham Open.

This is an open access article licensed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/)
which permits unrestricted, non-commercial use, distribution and reproduction in any medium, provided the work is properly cited.

L E
I C
R T
A
E D
C T
R A
E T
R