Proceedings of the 36th Chinese Control Conference

July 26-28, 2017, Dalian, China

Model Predictive Control for DC-DC Buck Power Converter-DC

Motor System with Uncertainties Using a GPI Observer
Han Wu, Lu Zhang, Jun Yang, Shihua Li
Key Laboratory of Measurement and Control of CSE, Ministry of Education, Nanjing 210096, Jiangsu
E-mail: [email protected]; [email protected]; [email protected]; [email protected]

Abstract: In this paper, angular velocity control of permanent magnet dc motor is addressed by using an output feedback
discrete-time model predictive control (MPC) algorithm. The operational constraints are imposed on the control input. The
control system is designed based on a topology of the dc-dc buck power converter-dc motor system in the presence of unknown,
exogenous, time-varying load torque disturbances. The generalized proportional integral (GPI) observer is used to estimate
the time-varying disturbances. Finally, the effectiveness of the proposed model predictive controller based on the generalized
proportional integral (GPI) observer algorithm is demonstrated by simulation results.
Key Words: Output Feedback, Model Predictive Control, Generalized Proportional Integral (GPI) Observer, DC-DC Buck
Power Converter-DC Motor System

1 Introduction of the torque disturbance acting on the shaft of the dc motor.

Different research works have proved its validity. For ex-
DC motor is widely used in industrial field due to its ad- ample, in [5], Linares-Flores, Reger and Sira-Ramirez apply
vantages of operation over a wide range. For example, a the algebraic approach to solve the problem of online torque
large number of dc motors are used in manipulators, mobile estimation in the boost-converter/dc-motor system. In [7],
robots, machine tools, railways, elevators and other industri- Grignion and Chen design three different estimators to esti-
al applications [1]. Thus, dc motors are vastly used in the mate the load torque disturbance.
occation where speed or torque require to be controlled. The ESO can estimate the uncertainties along with the s-
The earliest configuration of dc motor drives consists of tates of the system, enabling disturbance rejection or com-
a mesh of resistors connected in series or in parallel with pensation. Unlike traditional (linear or nonlinear) observers,
the dc motor. The advantages of the rheostatic control are the ESO estimates the effect of uncertainties, unmodeled dy-
its simple structure and low cost, however, rheostatic con- namics, and external disturbances acting on the system as an
trol suffers from poor efficiency, because a large amount of extended state of the original system [4]. In recent years, the
energy is dissipated as heat in the resistors. In addition, it generalized proportional integral observer also confirms its
cannot offer smooth regulation, resulting in abrupt changes efficiency in disturbance estimation and can be more accu-
and jerkiness. With the rapid development of power elec- rate. Compared with ESO, GPI uses more than one order
tronics, the rheostatic control of dc motor drives is obsolete. extended states in estimating total disturbance. It can be eas-
Until now, the dc chopper control for its advantages such as ily implemented because it does not need the exact knowl-
smaller size, lighter weight, higher efficiency and better con- edge of the system model and the observer design is based
trollability has been extensively used for dc motor drives [2]. on the linear configuration. There are some relative works
For instance, DC-DC power converters are widely used as s- in the converter-dc motor system, for example, in [6], the G-
mooth starters of dc motors. Moreover, there are different PI observer is used by Linares-Flores and Barahona-Avalos
kinds of dc-dc converters, such as buck converter [3], boost to estimate the exogenous torque disturbance, then the dis-
converter [5], buck-boost converter [6] as well as other ba- turbance is cancelled in an ADRC scheme. According to
sic circuits. The choice of converters depends on the desired the work presented by Sira-Ramirez and Oliver-Salazar in
operation and performance. [8], the GPI observer is used to estimate and eliminate the
The DC-DC buck converter is used as the starter of the lumped time-varying disturbance, the controller therein al-
permanent magnet dc motor to regulate angular velocity in so provides proper regulation even the control input gain is
this paper. It is generally known that the speed control al- imprecisely known. One similarity of the proposed meth-
gorithms of dc motors are cheaper and simpler than those ods in above mentioned articles lies in their observer and
of ac motors. Meanwhile, the speed of the dc motor can controller design are both based on the continuous system.
be regulated over a large range, both below and above rated However, it is well-known that for the control algorithm
speed can be easily achieved. The most challenging control in the microprocessor-based hardware, the continuous-time
problem is that it is necessary to maintain as accurately as controller should be converted to discrete time. Consequent-
possible of the speed even in the presence of torque distur- ly, we design a discrete GPI observer for the discrete model
bances posed on the shaft of the dc motor. In order to solve of the dc-dc buck power converter-dc motor system.
this problem, we need to have some knowledge of the torque Some other works on the dc-dc buck power converter-dc
disturbances [7]. However, due to the difficulties of direct motor system is referred to [3, 9–13], while the variable con-
measurement, for example, the cost of the torque sensor is straints are not explicitly taken into account therein. Model
very expensive, noises induced by measurement will bring predictive control (MPC) is known as a promising approach
some bad effects to the closed-loop system and so on. Esti- that the physical constraints can be handled effectively, both
mation has become a popular approach to measure the value

4906
on the input and/or state variables. It is actually a reced- where iL is the inductor current of the buck converter, vo is
ing horizon method in which the control input is calculat- the converter output voltage, ia is the dc motor armature cir-
ed through solving an optimization problem with physical cuit current, ω is the angular velocity of the motor shaft, ke
constraint over a finite number of future time steps at ev- is the counter electromotive force constant, km is the motor
ery time step [14]. In [15], a prediction-accuracy-enhanced torque constant, J0 is the moment of inertia of the rotor, b
continuous-time MPC with disturbance observer method is is the viscous friction coefficient of the motor, τL is the load
proposed for the disturbed systems. In [16], enhanced MPC torque. The duty ratio u(t) ∈ [0, 1] represents the control
with extended state observer (ESO) is adopted for control- signal.
ling the direct-drive elevators. As a result, the online MPC Moreover, the reference angular velocity is defined as
algorithm for the dc-dc buck power converter-dc motor sys- ω ∗ (t). It is desired to have the output angular velocity ω(t)
tem with discrete GPI observer is presented for the purpose asymptotically track the given reference trajectory ω ∗ (t), re-
of getting desired dynamics in closed loop. gardless of the effects of the unknown, but bounded load
In this paper, a discrete MPC controller with GPI observ- torque τL (t).
er algorithm is utilized to control the angular velocity of the 3 GPI Observer Design
dc motor. Firstly, a discrete linear generalized proportional
integral (GPI) observer is proposed to reconstruct the states To simplify the analysis without affecting its dynamics,
and estimate the lumped uncertainties of the considered sys- the system (1) is normalized by defining the following state
tem. Secondly, a discrete-time model predictive controller vector xT = (x1 , x2 , x3 , x4 ) = (iL , vo , ia , ω) , the follow-
(MPC) with input constraint is established based on the esti- ing normalized average system model (2) is obtained.
mations of the GPI observer.
1 E
This paper is organized as follows: Section 2 presents ẋ1 = − x2 + u,
the dynamic model of the combined dc-dc buck power L0 L0
1 1 1
converter-dc motor system. Section 3 presents the GPI ob- ẋ2 = x1 − x2 − x3 ,
server design and parameter regulation. Section 4 puts for- C0 C0 R0 C0
(2)
ward the design of the discrete-time model predictive con- 1 Ra ke
ẋ3 = x2 − x3 − x4 ,
troller subject to the control input constraint. Section 5 La La La
briefly describes the simulation paraments settings employed km b τL
ẋ4 = x3 − x4 − .
to test the effectiveness of the proposed control method. Sec- J0 J0 J0
tion 6 concludes the paper.
The load torque τL is considered as a disturbance, denoted
2 Dynamic Models of Buck Converter-DC Motor as d, then the state space formulation of the system can be
written as
ẋ = Ax + Bu u + Bd d,
 (3)



y = Cx.

where
 
  
⎡ ⎤


0 − L10 0 0
  
Z 1
 ⎢ − C01R0 − C10 0 ⎥
A=⎢
⎣
C0 ⎥,
0 1
−R − Lkea ⎦

a
La La
Fig. 1: Circuit of DC-DC Buck Power Converter-DC Motor 0 0 km
J0 − Jb0
⎡ ⎤ ⎡ ⎤
E
L0
0
Consider a dc-dc buck power converter acting as a smooth ⎢ 0 ⎥ ⎢ 0 ⎥
Bu = ⎢ ⎥ ⎢ ⎥
starter of permanent magnet dc brush motor in Fig. 1. The ⎣ 0 ⎦ , Bd = ⎣ 0 ⎦ ,
system has two components: a generic PWM-based DC-DC 0 − J10
buck converter and a permanent magnet dc brush motor with 
its armature circuit. The buck converter comprising a dc in- C= 0 0 0 1 .
put voltage source E, a PWM gate drive controlled switch A GPI observer is designed to estimate the lumped uncer-
V T , a diode V D, a filter inductor L0 , a filter capacitor C0 tainties. From the model in (3), the following formulas are
and a load resistor R0 . The permanent magnet dc motor deduced as
comprising a armature inductance La , a armature (or rotor)
winding resistance Ra , an induced electromotive force Ea . y =Cx,
The dynamic model is given as ẏ =CAx+CBd d,
diL0 ˙
ÿ =CA2 x+CABd d+CBd d,
L0 = −vo +uE, (4)
dt ˙
y (3) =CA3 x+CA2 Bd d+CABd d+CB ¨
d d,
dvo vo
C0 = iL0 − −ia , ˙
y (4) =CA4 x+CA3 Bd d+CA2 Bd d+CAB ¨
dd
dt R0
(1)
dia + CBd d(3) +CA3 Bu u.
La = vo −Ra ia −ke ω,
dt
dω Let z = [z1 , z2 , z3 , z4 ]T = [y, ẏ, ÿ, y (3) ]T be an 4-
J0 = km ia −bω −τL, dimensional vector and consider the following coordinate
dt

4907
transformation where ζi (i = 1, · · · , 7) are estimates of states zi (i =
⎡ ⎤−1 ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ 1, · · · , 7), respectively, ιi (i = 0, 1, · · · , 6) are the observ-
C 0 0 0
⎢ CA ⎥ ⎢ CBd ⎥ ⎢ er gains to be designed.
⎢ ⎥ ⎢ ⎥d− ⎢ 0 ⎥ ⎥d−
⎢
˙ ⎢ 0
⎥
⎥d¨ .
x=⎣ z− ⎣ ⎦ ⎣ (5) The estimation error equation is determined by
CA2 ⎦ CABd CBd ⎦ ⎣ 0 ⎦
CA3 CA2 Bd CABd CBd ⎡ ⎤
1−ι6 Ts
⎢ −ι5 1 Ts ⎥
The model in (3) can be re-arranged as follows: ⎢ ⎥
⎢ −ι4 0 1 Ts ⎥
⎢ ⎥
˙ d)
y (4) = f (y, ẏ, ÿ, y (3) , d, d, ¨ + mu, (6) e(k+1) = ⎢
⎢ −ι3 0 0 1 Ts ⎥ e(k),
⎥ (11)
⎢ −ι2 0 0 0 1 Ts ⎥
⎢ ⎥
⎣ −ι1 0 0 0 0 1 Ts ⎦
where m = CA3 Bu , then the system in (6) can be rewritten −ι0 0 0 0 0 0 1
as
ż = Am z + Bum u + Bf m f, the characteristic polynomial of the error system is
(7)
y = Cm z,
G0 (q) =(q−1)7 +ι6 (q−1)6 +Ts ι5 (q−1)5
where ⎡ ⎤
0 1 0 0 + Ts2 ι4 (q−1)4 +Ts3ι3 (q−1)3 (12)
⎢0 0 1 0⎥⎥, + Ts4 ι2 (q−1)2 +Ts5ι1 (q−1)+Ts6ι0 .
Am = ⎢
⎣0 0 0 1⎦
0 0 0 0 A seventh degree polynomials are chosen as the form
⎡ ⎤ ⎡ ⎤ ((q−1)2 +2ξwn +wn (q−1)+wn2 )3 ((q−1)+p). Comparing
0 0
⎢ 0 ⎥ with (12), we have
⎥
⎢ 0⎥⎥,
Bum = ⎢ ⎢
⎣ 0 ⎦ , Bf m = ⎣ 0⎦ ι6 = p + 6ξwn ,
m 1
 ι5 = (3wn2 + 12ξ 2 wn2 + 6ξwn p)/Ts ,
Cm = 1 0 0 0 .
ι4 = (12ξwn3 + 12ξ 2 wn2 p + 8ξ 3 wn3 + 3wn2 p)/Ts2 ,
By denoting z5 , z6 , z7 as f , f˙, f¨, respectively, the extend- ι3 = (8ξ 3 wn3 p + 12ξ 2 wn4 + 12ξwn3 0 + 3wn4 )/Ts3 ,
ed system can be written as
ι2 = (6ξwn5 + 12ξ 2 wn4 p + 3wn4 p)/Ts4 ,
⎡ ⎤ ⎡ ⎤
0 1 0 ι1 = (6ξwn5 p + wn6 )/Ts5 ,
⎢0 0 1 ⎥ ⎢ 0 ⎥
⎢
⎢0 0 0 1
⎥
⎥
⎢
⎢ 0 ⎥
⎥ ι0 = (wn6 p)/Ts5 ,
⎢ ⎥ ⎢ ⎥
z̄˙ = ⎢ ⎥ z̄+ ⎢ m ⎥u,
⎢0 0 0 0 1 ⎥ ⎢ ⎥ (8) where 0 < p < 2, ξ = 1, 0 < wn < 2.
⎢0 0 0 0 0 1 ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥ 4 MPC Controller Design
⎣0 0 0 0 0 0 1⎦ ⎣ 0 ⎦
0 0 0 0 0 0 0 0 The continuous time state space model for the dc-dc buck
converter-dc motor system given by (7) will be discretized
where z̄ = [z1 , z2 , z3 , z4 , z5 , z6 , z7 ]T . for the purpose of designing a discrete model predictive con-
The discretization of the extended system by using the ap- troller. Instead of using the approximation of differential e-
proximation of differential equations is presented as quations, here, the discretization of continuous-time mod-
el is performed together with the mechanism of zero-order
z̄(k + 1) = Gz̄(k) + Hu(k), (9) hold, which is commonly used in the computer controlled
systems. This discretization approach is presented in the fol-
where lowing formulas
⎡ ⎤ ⎡ ⎤
1 Ts 0 ⎧
⎨ Ā = e 
Am Ts
⎢ ⎥ ⎢ ⎥ ,
⎢ 0 1 Ts ⎥ ⎢ 0 ⎥ Ts Am τ
⎢ ⎥ ⎢ ⎥ B̄u = 0 e dτ · Bum , (13)
⎢ 0 0 1 Ts ⎥ ⎢ 0 ⎥ ⎩ T
G=⎢ 0 0 0 1 Ts ⎥ , H =⎢ mTs ⎥. B̄f = 0 s eAm τ dτ · Bf m ,
⎢ ⎥ ⎢ ⎥
⎢ 0 0 0 0 1 Ts ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥ where Ts is the sampling time, the discrete-time model is
⎣ 0 0 0 0 0 1 Ts ⎦ ⎣ 0 ⎦
presented as
0 0 0 0 0 0 1 0
z(k + 1) = Āz(k) + B̄u u(k) + B̄f f (k),
A discrete full-order GPI observer for the composite sys- (14)
y(k) = C̄z(k).
tem based on the knowledge of the angular velocity y(k) and
the control input u(k) is designed as Consequently, in the discrete model (14), matrices Ā, B̄u ,
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ B̄f and C̄ = Cm can be regarded as constant matrix. u(k)
ζ1 (k+1) ζ1 (k) ι6
⎢ ζ2 (k+1) ⎥ ⎢ ζ2 (k) ⎥ ⎢ ι5 ⎥ is designed for the discrete model subject to the input con-
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ζ3 (k+1) ⎥ ⎢ ζ3 (k) ⎥ ⎢ ι4 ⎥ straint.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ζ4 (k+1) ⎥ = G⎢ ζ4 (k) ⎥ +Hu(k)+ ⎢ ι3 ⎥ y(k)−ζ1 (k) , (10)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ To predict future dynamic of output based on the basic al-
⎢ ζ5 (k+1) ⎥ ⎢ ζ5 (k) ⎥ ⎢ ι2 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ gorithmic principle of predictive control theory and the dy-
⎣ ζ6 (k+1) ⎦ ⎣ ζ6 (k) ⎦ ⎣ ι1 ⎦
ζ7 (k+1) ζ7 (k) ι0 namic model (14), we define the prediction horizon Np , the

4908
control horizon Nc , and Nc ≤ Np . For the purpose of de- Optimization of Lagrange multipliers is commonly used
ducing the predictive equations of the system, we need the to directly search for the active constraints[18]. The param-
following assumptions [17]: eter vector U can be obtained by solving the following opti-
Assumption1: In addition to the control horizon, the mization problem:
amount of control input remains unchanged, namely, u(k +
[−2U Ep (k)+U Hp (k)U +2λ (M U −Γ)], (20)
T T T
i) = u(k + Nc ), i = Nc , Nc + 1, · · · , Np − 1. max
  min
Assumption2: The lumped disturbance remains un- λ≥0 U

changed after current time k, namely, f (k + i) = f (k),

the cost function (20) is minimized when
i = 1, 2, · · · , Np − 1.
Predictive equations of the system can be written as U ∗ = Hp−1 (Ep − M T λ), (21)
Y (k) = Ψz(k) + Φu U + Φf f (k), (15)
where the maximization problem (20) can be converted to
where the minimization problem by multiplying the objective func-
⎡ ⎤ ⎡ ⎤ tion by −1, where λ is the vector of Lagrange multipliers.
y(k+1|k) u(k) The cost function can be simplified as
⎢ y(k+2|k) ⎥ ⎢ u(k+1) ⎥
⎢ ⎥ ⎢ ⎥
Y (k) = ⎢ .. ⎥ , U (k) = ⎢ .. ⎥, Jλ = min −1 −1
[−2λ (M Hp Ep −Γ)+λ M Hp M λ],
T T T
⎣ . ⎦ ⎣ . ⎦ (22)
y(k+Np |k) u(k+Nc −1) λ≥0
⎡ ⎤
⎡ ⎤ C̄ B̄f and the terms which is independent of λ is neglected.
C̄ Ā ⎢ C̄ B̄f + C̄ ĀB̄f ⎥ According to the receding horizon optimization principe,
⎢ ⎥ ⎢ ⎥
⎢ C̄ Ā2 ⎥ ⎢ .. ⎥ the first element of the control sequence U ∗ (k) is imple-
Ψ=⎢ .. ⎥ f ⎢
, Φ = ⎢ . ⎥,
⎥
⎣ . ⎦ ⎢ N p−1 ⎥ mented, that is
⎣ ⎦
C̄ ĀNp C̄ Āi B̄f
u∗ (k) = [1, 0, · · · , 0]U ∗ (k). (23)
i=0
⎡ ⎤ 5 Simulation Results
C̄ B̄u 0 ··· 0
⎢ C̄ ĀB̄u C̄ B̄u ··· 0 ⎥
⎢ ⎥ To evaluate the proposed control method, detailed simu-
⎢ .. .. .. .. ⎥
Φu = ⎢
⎢ . . . . ⎥.
⎥ lation studies are performed by MATLAB/Simulink. The
⎢ 
Np ⎥ physical meanings and nominal values of the parameters are
⎣ ⎦
C̄ ĀNp−1 B̄u C̄ ĀNp−2 B̄u ··· C̄ Āi−Nc B̄u
listed in Table 1.
i=Nc

The predictive reference signals are given by Table 1: Parameters of DC-DC Buck Converter-DC Motor
 T Description Parameters Nominal Values
R(k) = r(k+1|k) r(k+2|k) · · · r(k+Np |k) , Input Voltage E 40(V )
Converter Inductance L0 8.9(mH)
and the cost function is designed by Converter Capacity C0 57.6(μF )
Load Resistance R0 492.6(Ω)
T Armature Inductance La 15.91(mH)
J = R(k) − Y (k) R(k) − Y (k) . (16) Armature Winding Resistance R0 6.14(Ω)
Viscous Friction Coefficient b 40.92 × 10−6 (N · m)
To formulate the optimization problem, the cost function Moment of Inertia of Rotor J0 7.95 × 10−6 (Kg · m2 )
is further calculated by plugging (15) into (16), then we can Constant of Motor ke k m 0.04913
deduce that

J = R − Ψz−Φf f
T
R−Ψz−Φf f To illustrate the robustness performance of the proposed
(17) MPC strategy, a classic PID controller is tested for compar-
−2U T ΦTu R−Ψz−Φf f +U T ΦTu Φu U, ison on the DC-DC buck converter-dc motor in the cases of
various sources of disturbances and uncertainties. The sim-
where the first term on the first line of equation (17) is inde-
ulation parameters are the same as shown in Table 1.
pendent on U . Thus we have
The tuning criterion of the controller parameters is to en-
 J ⇒ min
min sure a good balance among tracking, disturbance rejection
 Jy ,
U U and robustness performances. The controller parameters of
the proposed MPC law (23) are chosen to be
Jy = −2U T Ep (k)+U T Hp (k)U, (18)
Ts = 0.0001, Np = 86, Nc = 3,
where Ep = ΦTu R−Ψz−Φf f , Hp = ΦTu Φu .
Constraints of the control input can be re-formulated as while the observer parameters are set to be
follows:
M U ≤ Γ, (19) p = 0.2, ξ = 1, wn = 0.8.
where
    The PID controller parameters are
−1 0 · · · 0 0
M= , Γ= .
1 0···0 2×Nc
1 Kp = 0.0034, Ki = 0.135, Kd = 1 × 10−5 .

4909
 Fig. 2 shows the output angular velocity when the load the disturbances are 20 Hz and 0.05N ·m, respectively. Re-
torque τL changes from 0 to 0.05 N·m at t = 0.25s. Compar- sponse curves of the output angular velocity and duty ratio in
ison between PID and MPC-GPI shows that the robustness the presence of such a time-varying disturbance via the pro-
performance of the MPC-GPI strategy is much better. In posed MPC and PID control approaches are shown in Fig. 4
Fig. 2, when the load torque changes, the output angular ve- and Fig. 5, respectively.
locity using MPC departures rapidly from the reference sig-
nal; while PID takes about 0.1s for output velocity to track 500
the reference velocity again. The duty ratio of the switch can

Angular Velocity ω(t),rad/s

not be above 1 or below 0 due to the basic principle of PWM,
so a saturation function has been used to restrict the extent 400
of PID control quantity. As shown in Fig. 3, the dash-dotted 430
line represents the actual calculation value of PID control 300 415
law, the dotted and solid line represent the duty ratio curves 400
of PID and MPC-GPI methods respectively.
200 385
370
500 0.2 0.25 0.3 0.35 0.4
100
Angular Velocity ω(t),rad/s

400 PID MPC−GPI

0
420 0 0.1 0.2 0.3 0.4 0.5
300 Time,sec
400
380 Fig. 4: Angular velocity response curves in the presence of
200 time-varying disturbances
360
0.2 0.25 0.3 0.35 0.4
100
PID MPC−GPI 1.4
0
0 0.1 0.2 0.3 0.4 0.5 1.2
Time,sec
1
Duty Ratio u(t)

Fig. 2: Angular velocity response curves in the presence of

sudden load torque increase 0.8

0.6

1.4 0.4

1.2 0.2

1 0
Duty Ratio u(t)

0 0.1 0.2 0.3 0.4 0.5

0.8 Time,sec

0.6 Fig. 5: The actual calculation value of PID control law (dash-
dotted line), duty ratio curves of PID (dotted line) and MPC-
0.4 GPI (solid line) control strategy in the presence of time-
varying disturbances
0.2
It can be observed from the above two cases of simulation
0
0 0.1 0.2 0.3 0.4 0.5 results that although the PID controller could successfully
Time,sec eliminate the effects caused by load torque change, while
it fails to eliminate the effects caused by time-varying dis-
Fig. 3: The actual calculation value of PID control law (dash- turbances. As shown in Fig. 2-5, the proposed MPC sub-
dotted line), duty ratio curves of PID (dotted line) and MPC- stantially improves transient and static performance in the
GPI (solid line) control strategy in the presence of sudden presence of various disturbances and uncertainties including
load torque increase step and time-varying load torque changes as compared to
the PID approach.
We further investigate the robustness of the proposed M-
6 Conclusion
PC approach against time-varying disturbances. A generic
sawtooth waveform of time-varying load torque is taken to The work in this paper has addressed the senseless out-
acting on the motor system. The frequency and amplitude of put feedback discrete model predictive control (MPC) design

4910
with additional input constraint problem. A discrete GPI motor, in Proceedings of 1th International Conference on Con-
observer has been proposed for the involved lumped time- trol, Measurement and Instrumentation, 2016: 331–335.
varying disturbance estimation for dc-dc buck converter- [13] J.M. Alba-Martı́nez, R. Silva-Ortigoza, H. Taud, J. Alvarez-
dc motor system subject to load torque mutation and time- Cedillo, I. Rivera-Zárate, R. Bautista-Quintero, DC Motor
variation. It permits to take the sampling time into account, Speed Control via a DC to DC Buck Power Converter, in
Proceedings of 9th International Conference on Electronics,
overcoming the continuous MPC limitations and resulting in
Robotics and Automotive Mechanics, 2012: 288–293.
a good dynamic response, thus avoiding the effects of un-
[14] S. Kim, C. R. Park, J. Kim, Y. I. Lee, A Stabilizing Model
known load torque. The angular velocity of the dc motor is Predictive Controller for Voltage Regulation of a DC/DC Boost
regulated during each sampling time interval by solving the Converter, IEEE Trans. on Control Systems Technology, 22(5):
optimization problem. The best switching duty ratio which 2016–2023, 2014.
minimizes the cost function and satisfies the constraint is se- [15] J. Yang, W. X. Zheng, S. H. Li, B. Wu, Design of
lected and applied to the dc-dc buck converter-dc motor sys- a Prediction-Accuracy-Enhanced Continuous-Time MPC for
tem. The simulation results on the dc-dc buck converter-dc Disturbed Systems via a Disturbance Observer, IEEE Trans.
motor system have shown that the proposed method outper- on Industrial Electronics, 62(9): 5807–5816, 2015.
forms PID control approaches in the presence of various dis- [16] G. Wang, Y. Wang, J. Xu, N. Zhao, D. Xu, Weight-
turbances and uncertainties of the load torque. Note that the Transducerless Rollback Mitigation Adopting Enhanced M-
PC With Extended State Observer for Direct-Drive Elevators,
presented control approach is also expected to extend to oth-
IEEE Trans. on Power Electronics, 31(6): 4440–4451, 2016.
er DC-DC converters with mild modifications as the starter
[17] H. Chen, On port System and control systems, in Model Pre-
of the dc motor. dictive Control, H. Chen, Eds. Beijing: Science Press, 2013:
14–30.
References
[18] L. P. Wang, S. Chai, D. Yoo, L. Gan, K. Ng PID and Predic-
[1] G. G. Karady, K. E. Holbert, Electrical Energy Conversion and tive Control of Electrical Drives and Power Converters using
Transport: An Interactive Computer-Based Approach. New Matlab/Simulink. Singapore: Wiley-Verlag, 2015, chapter
Jersey: Wiley-Verlag, 2005, chapter 8. 8.
[2] K. T. Chau, Electric Vehicle Machines and Drives: Design,
Analysis and Application. Singapore: Wiley-Verlag, 2015,
chapter 2.
[3] R. Silva-Ortigoza, V. M. Hernández-Guzmán, M. Antonio-
Cruz, D. Muñoz-Carrillo, DC/DC Buck Power Converter as
a Smooth Starter for a DC Motor Based on a Hierarchical Con-
trol, IEEE Trans. on Power Electronics, 30(2): 1076–1084,
2015.
[4] S. E. Talole, J. P. Kolhe, S. B. Phadke, Extended-State-
Observer-Based Control of Flexible-Joint System With Ex-
perimental Validation, IEEE Trans. on Industrial Electronics,
57(4):1411–1419, 2010.
[5] J. Linares-Flores, J. Reger, H. Sira-Ramirez, Load Torque Esti-
mation and Passivity-Based Control of a Boost-Converter/DC-
Motor Combination, IEEE Trans. on Control Systems Technol-
ogy, 18(6): 1398–1405, 2010.
[6] J. Linares-Flores, J. L. Barahona-Avalos, H. Sira-Ramirez,
M.A. Contreras-Ordaz, Robust Passivity-Based Control of a
BuckCBoost-Converter/DC-Motor System: An Active Distur-
bance Rejection Approach, IEEE Trans. on Industry Applica-
tions, 48(6): 2362–2371, 2012.
[7] D. Grignion, X. Chen, Estimation of Load Disturbance Torque
for DC Motor Drive Systems Under Robustness and Sensitivity
Consideration, IEEE Trans. on Industrial Electronics, 61(2):
930–942, 2014.
[8] H. Sira-Ramirez, M. A. Oliver-Salazar, On the Robust Control
of Buck-Converter DC-Motor Combinations, IEEE Trans. on
Power Electronics, 28(8): 3912–3922, 2013.
[9] S. E. Lyshevski, Electromechanical Systems, Electric Ma-
chines, and Applied Mechatronics. USA: CRC-Verlag, 1999.
[10] R. Ortega, A. Loria, P. J. Nicklasson, H. Sira-Ramı́rez,
Passivity-Based Control of Euler-Lagrange Systems. London:
Springer-Verlag, 1998.
[11] A. Bhaumik, Y. Kumar, S. Srivastava, S. M. Islam, Perfor-
mance studies of a separately excited DC motor speed control
fed by a buck converter using optimized P I λ Dμ controller, in
2016 International Conference on Circuit, Power and Comput-
ing Technologies, 2016: 1–6.
[12] A. Chopade, S. Khubalkar, A. S. Junghare, M. V. Aware,
Fractional order speed controller for buck-converter fed DC

4911