Energy 131 (2017) 279e288

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Optimal vector control to a double-star induction motor

Imen Kortas*, Anis Sakly, Mohamed Faouzi Mimouni
Research Unit of Industrial Systems Study and Renewable Energy (ESIER), National Engineering School of Monastir (ENIM), Ibn El Jazzar, Skaness, 5019,
Monastir, University of Monastir, Tunisia

a r t i c l e i n f o a b s t r a c t

Article history: The problem of energy optimization of a Double Star Induction Motor (DSIM) using the concept of a Rotor
Received 28 September 2015 Field Oriented Control (RFOC) can be treated by an Optimal Control Strategy (OCS). Using OCS, a cost-to-
Received in revised form go function can be minimized and subjected to the motor dynamic equations and boundary constraints
7 March 2017
in order to find rotor flux optimal trajectories. This cost-to-go function consists of a linear combination of
Accepted 14 March 2017
Available online 20 March 2017
magnetic power, copper loss, and mechanical power. The Dynamic equations are represented by using a
reduced Blondel Park model of induction motor. From the Euler-Lagrange equation, a system of nonlinear
differential equations is obtained, and analytical solutions of these equations are achieved so as to obtain
Keywords:
Double star induction machine
a time-varying expression of a minimum-energy rotor flux. The current study discusses a saturation
Field-oriented control model with respect to the rotor flux, which has significant influence in the motor's parameters. A
Optimal control comparative study of simulation results given from conventional and optimized RFOC proves the pre-
Energy minimization sented strategy's validity and effectiveness.
Dynamic regime © 2017 Elsevier Ltd. All rights reserved.

1. Introduction nonlinear Euler-Lagrange equation from which an optimal flux

norm trajectory can be derived. The obtained equation was un-
Some of the largest opportunities to save energy and reduce solvable for an arbitrary torque. They proposed a suboptimal
operating costs in buildings and industrial facilities come from analytical solution aligned with a constant torque operation. In
optimizing electric motor systems. In general, the most part of Ref. [6], the authors present the solution of the nonlinear optimal
electricity assumed flux through motors mainly induction motors. control problem of three
phase induction motor (IM). A third or-
The DSIM is the dominant technology used today due to its high der nonlinear model described in arbitrary rotating frame of in-
performance, its high reliability, and its speed and torque capa- duction motor is used in this paper along with a quadratic
bilities. The conventional RFOC method operating at constant performance index. The problem is solved using the quasilinera-
rotor flux norm fixed at its standard level provides maximal effi- ization approach which converts the nonlinear optimal control
ciency when the system operates at its standard operating point problem into sequence linear quadratic optimal control problems.
[1]. Far from this, the machine's efficiency decreases; it can result In Ref. [7], a new minimum-time minimum-loss speed control al-
from a torque magnitude change. Thus other modes of flux gorithm for induction motors is suggested to obtain high perfor-
operation are required in order to reach system with optimal mance, as well as high efficiency, under field-oriented control with
performances. practical constraints on voltage and current. This algorithm utilizes
In this context; Lorenz et al. [2,3] have opted for an operating a two-stage control. In the transient stage, a maximum torque
loss modeled by copper and core losses. In order to compute the control algorithm is utilized to get the minimum-time response. In
minimal loss flux trajectory, they developed an optimal control the steady state, a minimum-loss control algorithm is applied to
strategy using a dynamic programming technique. C. Canudas de improve the efficiency. Simulation studies show the performance of
Wit et al. [4,5] considered a convex energy cost function including the proposed minimum-time minimum-loss control algorithm
the stored magnetic energy and coil losses. They developed a under field-oriented control.
S.M. Yang and F.C. Lin [8] propose a scheme that uses power
factor control with automatic measurement of the minimum-loss
power factor commands. A fuzzy logic compensator is included in
* Corresponding author.
the controller to improve the accuracy of the generated commands.
E-mail addresses: [email protected] (I. Kortas), [email protected]
(A. Sakly), [email protected] (M.F. Mimouni). The scheme is simple for implementation and does not require an a

http://dx.doi.org/10.1016/j.energy.2017.03.058
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
280 I. Kortas et al. / Energy 131 (2017) 279e288

priori knowledge of motor parameters. Experimental results have of three methods of LINAMP and TOPSIS and FUZZY. Finally,
validated the effectiveness of this scheme to minimize the motor sensitivity analysis and error analysis was performed for the
operating losses. A design of an adaptive nonlinear control system system.
for high performance induction motors is developed in Ref. [9]. The S. Hoseyn et al. [23] present an optimal design of a solar-driven
proposed control system is of the explicit model reference type. It heat engine Based on thermal and ecological criteria. In the pre-
consists of a nonlinear controller (inner loop) that controls the rotor sent investigation, thermodynamic analysis and an evolutionary
speed, an adaptation mechanism (outer loop) that involves a algorithm (EA) were employed to optimize the dimensionless
maximum likelihood estimator, communicating with a feedback ecological function, thermal efficiency, and dimensionless power
control law that uses the results of the adaptation mechanism to of a solar-driven engine system. Four scenarios were conducted
redesign the inner loop controller online. The advantage of syn- for optimization of the solar heat engine. In the first three, a
thesizing this type of controller lies in the fact that the desired traditional single objective optimization was employed separately
trajectory of the rotor speed is determined from the output of the with each objective function, regardless of other objectives. In the
reference model, while the control trajectories that lead to that fourth scenario, efficiency and power objectives were optimized
behavior are computed through the developed state feedback simultaneously using a nondominated sorting genetic algorithm
control law. The control system is simulated under a situation (GA) called the nondominated sorting genetic algorithm (NSGA-
where some of the parameters vary in the presence of noise. It is II). As in multiobjective optimization, a set of optimal solutions
shown that the adaptive controller keeps the performance of the called the Pareto optimal frontier was obtained instead of a single
drive system close to the desired performance even in the presence final optimal solution obtained in traditional single-objective
of uncertainty. The effect of measurement noise is also taken into optimization. Therefore, a process of decision making was
consideration to show that the controller is feasible for practical employed for selecting a final optimal solution. Three decision-
situations. making procedures were applied to find optimized solutions
M. H. Ahmadi et al. [20] present a developed ecological function from the Pareto optimal solutions in the objectives' space. The
for absorption refrigerators with four-temperature-level. Moreover, results obtained from four optimization scenarios were compared
aforementioned absorption refrigerator is optimized by imple- and discussed using a deviation index introduced in this paper. It
menting ecological function. With the aim of the first and second was shown that the optimal results obtained in single-objective
laws of thermodynamics, an equivalent system is initially deter- optimization with an ecological objective are very close to the
mined. To reach the addressed goal of this research, three objective corresponding results obtained in the multiobjective optimiza-
functions that the coefficient of performance (COP), the ecological tion, in that the power and thermal efficiency are optimized
function and thermoeconomic criterion have been involved in simultaneously.
optimization process simultaneously. Three objective functions are In this paper, minimum energy control will be solved by an
maximized at the same time. Developed multi objective evolu- alternative approach using Euler-Lagrange equation. By imposing
tionary approaches (MOEAs) on the basis of Non-dominated Sort- a transient mode to the DSIM drive, the nonlinear Euler equation
ing Genetic Algorithm (NSGA-II) method are implemented was successfully solved in an analytical form and offered a time-
throughout this work. varying expression of minimum energy rotor flux. This analyt-
In Ref. [21], authors propose an ecological and thermal approach ical solution was implemented in an optimal RFOC in which both a
for the Ericsson cryogenic refrigerator. Three objective functions deadbeat rotor flux controller and a saturation model are
(input power, coefficient of performance and ecological objective introduced.
function) are gained for the suggested system. Throughout the This paper is organized as follows: Section 2 is mainly intended
current research, an evolutionary algorithm (EA) and thermody- to describe the full DSIM model. The control law is detailed in
namic analysis are employed to specify optimum values of the input Section 3. In Section 4 we present the dynamic model of the DSIM.
power, coefficient of performance and ecological objective function In the fourth section the energy model of the DSIM is explained. In
of an Ericsson cryogenic refrigerator system. Four setups are Section 6, we develop the optimal control strategy by the energy-
assessed for optimization of the Ericsson cryogenic refrigerator. power cost function. In order to cheek the validity and the effec-
Throughout the three scenarios, a conventional single-objective tiveness of this proposed OCS using minimum-energy rotor flux
optimization has been utilized distinctly with each objective trajectory, we present simulation results by comparing the opti-
function, nonetheless of other objectives. Throughout the last mized RFOC with the conventional one.
setting, input power, coefficient of performance and ecological
function objectives are optimized concurrently employing a non-
dominated sorting genetic algorithm (GA) named the non- 2. Double-star induction motor model
dominated sorting genetic algorithm (NSGA-II). As in multi-
objective optimization, an assortment of optimum results named The DSIM is equipped with two symmetrical 3-phase armature
the Pareto optimum frontiers are gained rather than a single ulti- winding systems, electrically displaced by 30 . The state equation
mate optimum result gained via conventional single-objective of the motor model in the stationary reference frame fixed at the
optimization. Thus, a process of decision making has been uti- first stator, is described by (1).
lized for choosing an ultimate optimum result.
8 2 3 2 3 2 3
In Ref. [22], an optimization investigation of an irreversible >
>
Is1 " # Is1 Vs1
absorption heat pump system on the basis of a new thermo- >
> d6 7 A11 A12 6 7 6 7
>
> 6 7 6 Is2 7 þ B6 Vs2 7
ecological criterion is developed. The objective functions which >
>
> dt 4 Is2 5 ¼ 4 5 4 5
>
< A21 A22
considered are the specific heating load, coefficient of perfor- 4 r 4r 0
(1)
mance (COP) and the ecological coefficient of performance >
> dU
>
> 1
(ECOP). Three objective functions of the ECOP, COP and the >
> ¼ ðY
Cr Þ
> dt
> Jm
specific heating load are optimized simultaneously using the >
>
:
multi-objective optimization algorithm NSGAII. COP and ECOP Y ¼ a8 ðis1 þ is2 ÞT Jfr
are maximized and specific heating load is minimized in order to
get the best performance. Decision making is done by means where we have defined:
 I. Kortas et al. / Energy 131 (2017) 279e288 281

2 3 2 3 Fig. 1 shows the vector diagram defined in Equation (4), which is

a1 0 a2 0 a3 a4 refferred to the rotoreflux frame:
60 a1 0 a2 7 6
a4 a3 7
A11 ¼6
4 a2
7 A12 ¼6 7
0 a1 0 5 4 a3 a4 5
0 a2 0 a1
a4 a3
4. Dynamic model of the double star induction machine
   
a5 0 a5 0 a6
a7 In a reference frame rotating to the synchronous speed, equa-
A21 ¼ A22 ¼
0 a5 0 a5 a7 a6 tions defining the full-order dynamic model of the double star
asynchronous machine are given by (5) [4,13]:
2 3
b1 0 b2 0   8
6 0
6 b1 0 b2 7
7 J ¼ 0
1 >
>
> I_sðd;qÞ ¼
ðgI þ ðr_ þ pUÞJÞIsðd;qÞ
B ¼ 4 5 >
>
b2 0 b1 0 1 0 <
4_ rðd;qÞ ¼
ðaI þ r_ JÞ4rðd;qÞ þ bI
0 b2 0 b1 (5)
>
>
>
> _ ¼
Kl þ Y
>
:U
ai, i ¼ 1,2,…,8 and bj, j ¼ 1,2 are given in Appendix A.Cr is the load Jm Jm
torque,Y is the electromagnetic torque,Jm is the total moment of
     
inertia of the rotor, Is1d þ Is2d Isd Vsd1 þ Vsd2
with Isðd;qÞ ¼ ¼ ; Vsðd;qÞ ¼
In the current-controlled PWM drive, the control laws would Is1q þ Is2q Isq Vsq1 þ Vsq2
force the stators currents to be exactly equal to the reference cur-        
Vsd 4rd 2
M2 ;
rents. Under these conditions it is reasonable to assume that the ¼
Vsq
; 4rðd;qÞ ¼
4rq
; s 1 ¼ 1
M
Lr Ls s2 ¼ 1
Lr Lsp
rotor dynamics is described by the reduced model as follows: !    
2 1 0 0 1
8 " # g ¼ s1 Ls þ1s2 Lsp Rs þ ML2 Rr ; I ¼ ; J¼ ; a ¼ RLrr ;
Is1 0 1
1 0
>
> dfr r

>
> ¼ A22 fr þ A21
>
> dt Is2 b ¼ aM
>
< By noting us ¼ r_ is the sleep frequency, Rs and Rr are stator and
> dU 1 (2) rotor resistances, Ls and Lr are stator and rotor inductances; M is
>
> ¼ ðY
Cr Þ
>
> dt Jm the magnetizing inductance, Lsp is the principal cyclic inductance,
>
>
: Kl is the load torque constant. Isd1 , Isd2 , Isq1 and Isq2 are respectively
Y ¼ a8 ðIs1 þ Is2 ÞT Jfr
the direct and quadrature current of stator 1 and stator 2.
Vsd1 ; Vsd2 ; Vsq1 ; Vsq2 are respectively, the stator voltages in d-q axis
of each stator, p is the poles number,
3. The proposed control strategy The electromagnetic torque expressed as follows:

The magnetizing current vector imr of the rotor flux is defined Y ¼ Cr $ðIs1 þ Is2 Þtðd;qÞ $j$4rðd;qÞ (6)
by:

1
imr ¼ is1 þ is2 þ ir (3)
a0
5. Double-star induction motor reduced model
We suppose that the first star is used to control the rotor flux
and the second star is used to control the torque. Under these hy-
In order to eliminate the nonlinear terms in the DSIM dynamic
potheses, Equation (3) can be written as:
model and allow an easy resolution of a possible optimal control
8 problem, the system can be forced into a current-command mode
< is1 ¼ imr
>
using high-gain feedback. To achieve this goal, a PI current loop can
1 (4)
>
: is2 ¼
ir be applied [1]. Furthermore, the reduced order current-fed DSIM
a0 model can be simplified by restraining the PI to a sample propor-
tional gain [7,16], as follows:

s1 Ls þ s2 Lsp  
Vsðd;qÞ ¼ U
Isðd;qÞ (7)
ε

with
   
u1 Is1d þ Is2d
U¼ ¼
u2 Is1q þ Is2q

which 0 < ε < 1 and U is a new command of the system.

The standard reduced model form is obtained through singular
perturbation. It can be defined as:
8
> 4_ ¼
ðaI þ r_ JÞ4rðd;qÞ þ bU
< rðd;qÞ
t (8)
>U
: _ ¼
Kl þ cU J4rðd;qÞ
Jm Jm

Fig. 1. Diagram of currents vectors. where:

282 I. Kortas et al. / Energy 131 (2017) 279e288

3 M Pm ¼ UY (17)
c¼ p
2 Lr
In term of rotor variables and torque current ðIsq Þ, we get:

M
Pm ¼ 3 2 =
4 Isq U (18)
Lr r
6. Energy model of the DSIM

The instantaneous active power of the DSIM in the d-q rotating

7. Minimum energy approach
frame is given by:
 t 7.1. Energy-power cost function
Pa ¼ 3 2 Vsðd;qÞ Isðd;qÞ
=
(9)
An optimal control consists on minimization of a cost function.
with “t” denote the transpose of the considered vector. In this study, the cost function can be defined as the integral of an
From system (5), the input power is given by: index f ðIsd ; Isq ; 4r ; UÞ given as follows [10e12]:
   
  t  t 
ZT
Pa ¼ 3 2 s1 Ls þ s2 Lsp : I_sðd;qÞ
=
Isðd;qÞ þ g$ I_sðd;qÞ Isðd;qÞ 
J¼f Isd ; Isq ; 4r ; U dt (19)
 t    t 0

2ha 4rðd;qÞ Isðd;qÞ þ hpU 4rðd;qÞ JIsðd;qÞ
This index corresponds to the weighted sum:
(10) 
f Isd ; Isq ; 4r ; U ¼ b1 WL þ b2 PJ þ b3 Pm (20)
The relation between stator and rotor currents can be given as
follows: The weighting factors b1 , b2 and b3 are used to scale power-
energy combined convex criteria terms defined above. Minimize
1 
the cost function provides two important advantages: first, mini-
Irðd;qÞ ¼ 4rðd;qÞ
Msr Isðd;qÞ (11)
Lr mizing the corresponding magnetic energy stored, allowing thus
This yields the following input active power expression: maximizing the power factor, the second being the minimization of
losses in the windings thus increasing the machine efficiency.
 
  t  Using Equations (14), (16), (18) and (20), the cost function is
Pa ¼ 3 2 s1 Ls þ s2 Lsp $ I_sðd;qÞ
=
Isðd;qÞ
given as follows:
 
1  t  2 
   1
 3
þ $ 4_ rðd;qÞ 4rðd;qÞ
Lr 6 b1 s1 Ls þ s2 Lsp $ u21 þ u22 þ 7 $42
2Lr r
  6 7
t     t   ZT 6   ! !7

3 2 Rs I_sðd;qÞ þ Rr I_rðd;qÞ þ UY 6 2   7
Isðd;qÞ Irðd;qÞ 6 M Rr 7
=
J ¼ 32 =
6 þb2 Rs þ Rr u21 þ u22
2 42r 7dt
6 Lr L 7
0 6 7
r
(12) 6   7
4 b d M 5
On the other hand, the input active power can be expressed by:
2
42r þ b3 p 4r u2 U
Lr dt Lr
v (21)
Pa ¼ W þ PJ þ Pm (13)
vt Z T 
  b d 2 b
4r The integral
2 ð4r Þ dt ¼ 2 ð42r ð0Þ
42r ðTÞÞ, has no
By means of a field-oriented control drive, 4rðd;qÞ ¼ , the 0 L r dt Lr
0 effect in the optimizing of the problem and can be omitted from the
derivate of stored magnetic given in (13) can be expressed from (12)  
u1
as follows: integral. Considering the new control vector U ¼ , the system
u2
     described by (8) can be defined as follows:
v s1 Ls þ s2 Lsp 1 2
W ¼ 32 =
$ u21 þ u22 þ 4r (14) 8
vt 2 2Lr
< 4_ r ¼ a4r þ bu1
>
(22)
From Equation (12), we can also define the total copper loss as:
>
:U _ ¼
Kl U þ cu2 4r
  t     t   Jm Jm
PJ ¼ 3 2 Rs I_sðd;qÞ
=
Isðd;qÞ þ Rr I_rðd;qÞ Irðd;qÞ The optimal control problem will be defined as follows:
(15) Finding an optimal control ðu*1 Þ and ðu2 Þ that transfers the dy-
namic system described in (22) from the initial states: 4r ð0Þ ¼ 4r0
By using system (5) and (15), those losses can be expressed with and Uð0Þ to admissible final states 4r ðTÞ and UðTÞ, so that the cost
respect to U and 4r as follows: function:
 2 !   3R
!
M 3 d42r ZT 
b2   
2 2 r 2
PJ ¼ 32
=
Rs þ Rr $ u1 þ u2
4

Lr 2 L2r r 2Lr dt Jr ¼ 42r ð0Þ
42r ðTÞ þ r1 u21 þ r2 u22 þ q1 42r þ q2 4r u2 U dt
Lr
(16) 0
(23)
Finally, from Equation (12), the mechanical power transmitted
to the rotor is given by:
 I. Kortas et al. / Energy 131 (2017) 279e288 283

is minimized:where the weighting factors r1 , r2 , q1 and q2 must be

€ ¼ 2 _ U
positives. U 4_ U þ a4 4_ r þ a5 U þ a6 42r (30)
4r r 4r
 2 !
 M
r1 ¼ 3 4 s1 Ls þ s2 Lsp 41 þ 3 2
= =
Rs þ Rr 42 ; r2 with
Lr
¼ r1 þ r0 a3 l2 l l l l l l
a0 ¼ ; a ¼ 4; a ¼ 6; a ¼ 7; a ¼ 4; a ¼ 7; a ¼ 8
l1 1 l1 2 l1 3 l1 4 l2 5 l2 6 l2
a1 Rr 3M
q1 ¼ 3 4 =

32 =
a2 þ a3 q0 ; q2 ¼ a
Lr L2r 2 Lr 4
The optimal control problem amounts to determining the
optimized values of ðu*1 Þ and ðu2 Þ witch are defined as follows: 7.2.1. Linear time-varying motor speed
In this paper, the mechanical process operations are restricted to
8
>
>  1  
two modes in order to limit the study to the transient operation.
< u1 ¼ a7 4r þ a7 4r
> The first mode is the acceleration of the rotor up to given velocity.
  (24) The second mode is when the motor speed is constant.
>
>
> J
: u2 ¼
m U _  þ Kl
f U Then the motor speed can be chosen as the following
c4r Jm expression:
By injecting the expressions of u1 and u2 from system (24) in
U ¼ c0 t þ c1 (31)
Equation (13) we get:
0 with c0 > 0
ZT 2   
U_ l4 l7 Hence the condition given by (30) has no physical meaning and
Jr ¼ @ l1 4_ 2r þ l2 þ l3 4_ r 4r þ þ l5 UU_ þ l6 42r þ
42r 42r 42r can be skipped. By substituting the expression of U by U* given in
0
1 (31), Equation (29) becomes:
  
2A
þ l8 U dt € r 43r
a2 44r ¼
g1 t 2 þ g2 t þ g3
4 (32)

(25) where:

with l1 ¼ a2rM1 2 ; l2 ¼ r2cJ2m ; l3 ¼ aM

2
2r2
2; l4 ¼ 2rc22Jm k; l5 ¼ qc2 Jm ; g1 ¼ a3 c20 ; g2 ¼ a1 c20 þ 2a3 c0 c1 ; g3 ¼ a0 c20 þ a1 c0 c1 þ a3 c21
l6 ¼ Mr12 þ q1 ; l7 ¼ cr22 k2l ; l8 ¼ qc2 kl ;li:1 a 8 must be positives con- Despite the extreme difficulty to solve this Equation (31), we
stants defined accordingly to the condition on the constants r1 ; q1 tried to finding a solution using a mathematical calculation; so, we
and q2 . applied the method of integration by parts to this equation, during
a time interval ½0; t.
By integrating (31), we generate the first order differential
7.2. Development of Euler-Lagrange equation equation:

3 3 a2 5 3 3 a2 5
This integral given by (25) can be expressed as follows [18,19]: _
4ðtÞ4r ðtÞ
4r ðtÞ
_
4 ðtÞ
4ð0Þ4r ð0Þ þ 4r ð0Þ þ 4 ð0Þ
5 r 5 r
 
ZT   g t 3 g t 2
¼
1 þ 2 þ g3 t
Jr ¼ _ ; U dt
L 4_ r ; 4r ; U (26) 3 2
0 (33)
If we make a second integration, we have:
and can be solved using Euler-Lagrange equation [14,15,17], with
respect the follows condition: This integral has a minimum value 4_ *r Zt 
a2 5
and U* if their trajectory satisfies the following conditions: _
4ðtÞ43 3
r ðtÞ
4r ðtÞ
_
4 ðtÞ
4ð0Þ4 3 3
r ð0Þ þ 4r ð0Þ
5 r
v     
_ ; U
v v L 4_ ; 4 ; U
 0
L 4_ r ; 4r ; U r r
_ ;U ¼0 (27) a2 5 
v4r vt v4_ r þ 4 ð0Þ dt
5 r
and Zt   
g1 t 3 g2 t 2
   ¼
þ þ g3 t dt (34)
v     3 2
_ ;U
v
L 4_ r ; 4r ; U
v _ ;U
L 4_ r ; 4r ; U ¼0 (28) 0
vU vt vU_
This implies:
By using the expression of the cost function described in (25),
the previous conditions (27) and (28) becomes: a 46 ðtÞ a2 46r ð0Þ  a 

2 r þ
4_ r ð0Þ43r ð0Þ þ 43r ð0Þ þ 2 45r ð0Þ t
30 30 5
2  
U_ U _ U2 g1 t 4 g2 t 3 g3 t 2
€ r ¼
a0
4
a1 U þ a2 4r
a3 3 (29) ¼
þ þ þk (35)
43r 4r
3 4r 12 6 2

and where k is an appropriate positive constant.

284 I. Kortas et al. / Energy 131 (2017) 279e288

This yields: This equation gives a solution described by the following

expression:

30 a2 46r ð0Þ  a 
46r ðtÞ ¼
4_ r ð0Þ43r ð0Þ þ 43r ð0Þ þ 2 45r ð0Þ t
a2 30 5
   pffiffiffiffi
rffiffiffiffiffiffi pffiffiffiffi 
g1 t 4 g2 t 3 g3 t 2 g0
þ þ þ
k (36) sðtÞ ¼ 42r ðt0 Þe
2 a2 ðt
t0 Þ
þ 1
e
2 a2 ðt
t0 Þ
(46)
12 6 2 a2

Finally the optimal solution of the rotor flux is given by the Finally, the optimal rotor flux is given by:
following expression:

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
6 30 a2 46r ð0Þ  a 2
 g 1 t 4 g 2 t 3 g 3 t 2
4r ðtÞ ¼
4_ r ð0Þ43r ð0Þ þ 43r ð0Þ þ 45r ð0Þ t þ þ þ
k (37)
a2 30 5 12 6 2

7.2.2. Constant speed operation sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

rffiffiffiffiffiffi
By varying the proportionality constant Kl , the load torque pffiffiffiffi g0  pffiffiffiffi
4r ðtÞ ¼ 42r ðt0 Þe
2 a2 ðt
t0 Þ þ 1
e
2 a2 ðt
t0 Þ (47)
varies while maintaining a reference speed still constant. The motor a2
speed can be described by the following expression:

U ¼ c1 (38)
By substituting this expression for the rotor speed in Equation
7.3. The rotor flux Level's deadbeat control
(30), we obtain:
In order to have an appropriately-optimized RFOC, the motor
€ r 43r
a2 44r ¼
g0
4 (39)
parameter changes caused by the magnetic saturation effect should
be taken into account. The problem becomes difficult to resolve
where g0 ¼ c21 a3 by replacing 4_ r with q, we obtain:
when the saturation effect is taken into account in this offline OCS,
due to the calculus complexity level. But this effect can be easily
dq d4r dq dq
€ r ¼ q_ ¼
4 ¼ 4_ ¼ q (40) introduced in the RFOC strategy. In a rotating reference frame, the
d4r dt d4r r d4r
rotor flux can be expressed in terms of the flux current as follows:
then Equation (39) became: Msr
  4r ðtÞ ¼ I ðtÞ (48)
1 þ ðd=dtÞtr sðd;qÞ
g
qdq ¼ a2 4r
03 d4r (41)
4r Taking the saturation model into account in the RFOC strategy is
aimed at obtaining a good dynamic flux level tracking. Since the
whether field orientation control provides flux only in the d axis, it can be
assumed that saturation only occurs in the d-axis [22]. However,
g0
q2 þ q0 ¼ a2 4r þ (42) under steady state conditions magnetizing rotor current imr can be
42r confounded with isðd;qÞ . The experimental curve of the magnetizing
g0 rotor current imr versus rotor flux 4r ðtÞ can be expressed by the
with: q0 ¼ a2 42r ðt0 Þ þ
4_ r ðt0 Þ2 following polynomial function [18]:
4r ðt0 Þ2
The phase of constant speed operation that we have chosen is
obtained while having 4r ðtÞand 4_ r ðtÞ both positive. Equation (41) imr ¼ s1 4r þ s2 43r þ s3 45r (49)
can be expressed as follows:
It is obvious to observe that the Msr and tr parameters in the
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rotor flux Equation in (49) will be a function of the flux [22,24]. The
g0
q ¼ 4_ r ¼ ± a2 42r þ
q0 (43) mutual inductance in a steady state operation can be given by:
42r
4r 1
Mcsr ¼ ¼ (50)
after replacing q0 by its value, the real solution of Equation (41) is imr s1 þ s2 42r þ s3 44r
such that:
pffiffiffiffiffiffi with s1 ; s2 and s3 are the coefficients of the magnetizing curve given
pffiffiffiffiffi g0 in Fig. 2.
4_ r ðt0 Þ ¼ a2 4r ðt0 Þ þ (44)
4r ðt0 Þ Using the least squares method, the values of those coefficients
can be deduced: s1 ¼ 38.1; s2 ¼
36; s3 ¼ 1022.1in the case of a
by making the following change of variable: sðtÞ ¼ 42r , a differential reduced model not considering iron loss, the reference value of
first order equation can be obtained from Equation (44): *
current Isðd;qÞ ðnTs Þ can be written with respect to the reference rotor
pffiffiffiffiffi pffiffiffiffiffiffi flux 4*r ðnTs Þ, tr ðnTs Þ and the mutual inductance Msr ðnTs Þ as follows
s_ ðtÞ ¼
2 a2 sðtÞ þ 2 g0 (45)
[23,25]:
 I. Kortas et al. / Energy 131 (2017) 279e288 285

8 8
> Isd > Isq
>
< Isd1 ¼ >
< Isq1 ¼
2 2
(52)
>
> I >
> I
:I sd :I sq
sd2 ¼ sq2 ¼
2 2

 Case n 2: we inject a current Isd in the first star and a quadrature

current Isq in the second star. We control the flux by a star and
the torque by the other. Then we have:
 
Isd1 ¼ Isd Isq1 ¼ 0
(53)
Isd2 ¼ 0 Isq2 ¼ Isq

 Case n 3: the star 1 provides a constant rotor flux equal to

0.315Wb and the star 2 insures the production of the torque and
compensates the lack of power.
Fig. 2. Mutual inductance versus rotor flux.
 
Isd1 ¼ Isd Isq1 ¼ 0
(54)
Isd2 ¼ 0 Isq2 ¼ Isq
IsqL
" ∧
# pffiffiffiffiffiffiffiffi
1 4*r ðnTs Þ 4r ðnTs Þ with IsqL ¼ 2 : IsN
*
Isðd;qÞ ðnTs Þ ¼

1
expð
Ts =tr ðnTs ÞÞ Msr ðnTs Þ Msr ððn
1ÞTs Þ The DSIM models with conventional and optimal RFOC were
" numerically simulated. The conventional RFOC is based on a con-
expð
Ts =tr ððn
1ÞTs Þ 4*r ððn
1ÞTs Þ stant value-rated flux trajectory. For each case studied, a saturation


1
expð
Ts =tr ððn
1ÞTs Þ Msr ððn
1ÞTs Þ model with respect to the rotor flux was taken into account.
∧
# In order to be restricted to the transient operation, only the
4 ððn
1ÞTs Þ acceleration (from t ¼ 0s to t ¼ 1s) and permanent (from t ¼ 1s to

r þ i*sðd;qÞ ððn
1ÞTs Þ
Msr ððn
2ÞTs Þ t ¼ 2s) modes were taken into account in the motor speed profile,
as shown in Fig. 4.
(51)
Fig. 5 present the minimum energy flux trajectory correspond-
ing to the motor speed profile, all the unknown boundary condi-
with Ts is the settling time for the deadbeat control, 4*r is the rotor
∧ tions and parameters shown in the optimal rotor flux expression in
flux reference, and 4r as the estimated one. Equations (37) and (47) were determined.
Fig. 6 and Fig. 7 indicate that major energy-loss savings with
dynamically optimized rotor flux trajectories. It is obvious to
observe that the RFOC taking the magnetic effect into account and
7.4. The RFOC scheme
operating with optimal flux leads to the best DSIM energy con-
sumption decreasing. This decrease is more considerable for smalls
The proposed control scheme was studied by simulation as
loads.
shown in Fig. 3. A time varying flux is predicted by using the
To better express the difference between the three cases, energy
expression in Equations (37) and (47) with respect to the motor
saving of the optimum rotor flux trajectories is demonstrated by
speed reference injected in the input of the deadbeat rotor flux
the following comparison with the constant value-rated flux tra-
controller.
jectory summed up in Table 1 and Table 2.
Referring to simulation results we deduce that:

8. Simulations results - It is obvious to observe that the RFOC operating with optimal
flux leads to the best DSIM energy consumption decreasing. This
The DSIM models with conventional and optimal RFOC were decrease is more considerable for small loads.
numerically simulated. The conventional RFOC is based on a con- - Case 1 and 2 seems the better efficient in terms of energy saving.
stant value-rated flux trajectory. In both of these two control laws, a - We also observe a great reduction of energy in case 3, especially
saturation model with respect to the rotor flux was taken into for large loads.
account.
The motor parameters are Rs ¼ 0:4U, Rr ¼ 0:096 U, Lr ¼ 8:9mH
and Ls ¼ 81:2mH. The mechanical parameter are Jm ¼ 0:6 and
Kl ¼ 0:7. 9. Behavior of the DSIM during an overload operating
The load torque is assumed to be proportional to the reference
speed with Kl ¼ 0:7 is the constant of proportionality. Study is We want to evaluate the influence of the variation of speed on
carried out over a period of 2 s. the proposed minimum-energy method, so we proceeded to a ve-
The simulations are established with MATLAB software and locity profile described by Fig. 8.
have presented three cases:
- 0se0.1s: transient regime, the machine switches to zero speed
 Case n 1: we make a symmetric control, so we introduce the at nominal rotor speed (150rd/s).
same current in each stator. - 0.1se0.5s: operation at a constant rotor speed (150 rad/s).
286 I. Kortas et al. / Energy 131 (2017) 279e288

Fig. 3. RFOC scheme of the DSIM.

Fig. 4. Reference rotor speed.

- 0.5se1s: over rated rotor speed operation: the machine passes

Fig. 5. Optimal rotor flux.
from 150 rad/s to 180rd/s.
- 1.5s to 2: permanent regime: the machine operates at a constant
rotor speed (180 rad/s). 10. Conclusion

Fig. 9 and Fig. 10 present respectively a consumed energy and This paper presents an optimal control strategy using energy
losses in both conventional and optimized RFOCs. minimization in transient regime. The criterion used in this OCS is
The simulation result confirms the decrease of the consumed subjected to dynamic equations of a DSIM reduced-order model.
energy in both nominal and overloads operating of the DSIM due to The rotor flux and the motor speed are considered as state vari-
the reduction of losses. This reduce is more considerable at the ables. The optimal control problem is then treated by the Euler-
rated value of motor speed. Lagrange equation and a minimum energy rotor flux solution is
Energy saving of the optimum rotor flux trajectories is demon- successfully determined in an analytical form. This time-varying
strated by the following comparison with the constant value-rated solution is implemented in an RFOC in which a deadbeat rotor
flux trajectory summed up in Table 3. flux controller and a saturated model with respect to the rotor flux
 I. Kortas et al. / Energy 131 (2017) 279e288 287

Fig. 8. Profile of reference rotor speed.

Fig. 6. Energy in both conventional and optimal control for three cases for small loads.

Fig. 9. Energy under conventional and optimal control.

Fig. 7. Energy in both conventional and optimal control for three cases for large loads.

Table 1
Gain of energy for small loads.

Time (s) 0.5 0.52 0.54 0.56 0.6 0.8 0.9

Gain of energy (J) ecase1-
25.75 114.36 266.97 327.55 402.05 656.92 721.52
Gain of energy (J) ecase2-
26.08 111.09 266.72 327.86 402.36 657.13 721.70
Gain of energy (J) ecase3-
27.65 99.65 254.40 305.96 390.33 649.67 718.17

Table 2
Gain of energy for large loads.

Temps (s) 1 1.2 1.4 1.6 1.7 1.8 2

Gain of energy (J) ecase1- 752.86 734.27 642.95 534.30 492.14 464.68 454.97
Gain of energy (J) ecase2- 753.02 734.42 643.09 551.35 513.46 465.92 402.39
Gain of energy (J) ecase3- 765.45 746.81 654.45 568.75 556.57 486.08 495.32
288 I. Kortas et al. / Energy 131 (2017) 279e288

a0 a ur M 1
a3 ¼ ;a ¼ 0 ;a ¼ ;a ¼

Tr ðL1 þ L2 Þ 4 L1 þ L2 5 Tr 6 Tr

3np M L L
a7 ¼ ur ; a8 ¼ ; b1 ¼ 2 1 2 ; b2 ¼ 2 2 2
2Lr L1
L2 L1
L2

References

[1] Mimouni MF, Dhifaoui R. Modelling and simulation of double-star induction

machine vector control using copper losses minimization and parameters
estimation. Int J Adapt Control Signal Process 2002.
[2] Lorenz RD, Yang SM. Efficiency-optimized flux trajectories for closed cycle
operation of field oriented induction machine drives. IEEE Ind Appl Soc Annu
Meet 1988;1:457e762.
[3] Chiasson John. Modelling and high-performance control of electric machines.
IEEE press series on power engineering. Hobokon, New Jersey: John Willey &
Fig. 10. Losses under conventional and optimal control. sons, Inc; 2005.
[4] Canudas de Wit C, Ramirez J. Optimal torque control for current-feeded in-
duction motors. IEEE Trans Autom Control 1999:44.
[5] Didier G, de Wit C, Ramirez J. Nonlinear H2 and H∞ optimal controllers for
Table 3 current-fed induction motors. IEEE Trans Autom Control 1999:44.
Compared energy consumption of 20 KW DSIM under rotor flux oriented controls [6] Jasem MKh, Jaddu HM. The optimal control of a three-phase induction motor.
for different values of rotor speed. In: Proceedings of the 25th IASTED conference, lanzarote, Spain; 2006.
[7] Chang JH, Kim BK. Minimum-time minimum-loss speed control of induction
motors under field-oriented control. IEEE Trans Ind Electron 1997:44.
[8] Yang SM, Lin FC. Loss-minimization control of vector-controlled induction
motor drives. J Chin Inst Eng 2003;26:37e45.
[9] Rubaai A, Uribina O, Kankam MD. Design of an adaptative nonlinear
controller-based optimal control theory for a voltage source induction motor
drive system. Industry applications conference. In: Thirty-sixth IAS annual
meeting conference record of the IEEEvol. 2; 2001. p. 1279e84.
[10] Geering HP. Optimal Control with engineering application. Berlin Heidelberg:
Springer-Verlog; 2007.
[11] Ramirez JM. Contribution  a la commande optimale des Machines Asyn-
chrones. Frence: Automatic Laboratory of Grenoble; 1998.
[12] Chiasson J. Modelling and high-performance control of electric machines. IEEE
press series on power engineering. Hobokon, New Jersey: John Willey & sons,
Inc; 2005.
[13] Mimouni MF, Dhifaoui R. A robust air-gap flux estimation for speed sensorless
are introduced. Saturation effects are modeled in the RFOC to vector control of double-star induction machine. 2004.
provide a state-dependent constraint on the rotor flux magnitude [14] Weinstock R. Calculus of variations with applications to physics and engi-
neering. New York: Department of Mathematics, Stanford University, Dover
and produce a good tracking of the dynamically rotor flux trajectory Publication, Inc; 1974.
by using the deadbeat controller. The load is treated according to [15] Russak. Calculus of variations. Monterey, California: Department of Mathe-
motor speed, which is appropriate for many mechanical system matics, Naval Postgraduate School; 2002.
[16] Abdelati R, Mimouni MF. Analytical solutions of optimized energy con-
applications. The analysis of this optimal control law is accom- sumption of induction motor operating in transient regime. Eur J Control
plished by comparing it with a conventional RFOC. The latter 2011:17.
allowed much lower energy consumption and a better efficiency [17] Locatelli A. Optimal control. Birkhauser Verlog; 2001.
[18] Buoro D, et al. Optimization of a distributed cogeneration system with solar
than the conventional RFOC. district heating. Appl Energy 2014;124:298e308.
[19] MacArthur JW, et al. Application of numerical methods for predicting energy
transport in earth contact systems. Appl Energy 1983;13:121e56.
[20] Ahmadi MH, Ahmadi MA, Mehrpooya M, Hosseinzade H, Feidt M. Thermo-
dynamic and thermoeconomic analysis and optimization of performance of
Appendix A irreversible four-temperature-level absorption refrigeration. Energy Convers
Manag 2014;88:105e9.
[21] Ahmadi MH, Ahmadi MA. Thermodynamic analysis and optimization of an
Coefficients: irreversible Ericsson cryogenic refrigerator cycle. Energy Convers Manag
2015;89:147e55.
M2 M2 M [22] Ahmadi MH, Ahmadi MA, Mehrpooya M, Sameti M. Thermo-ecological anal-
L1 ¼ Ls
; L2 ¼ Lps
;a ¼ ; ysis and optimization performance of an irreversible three-heat-source ab-
Lr Lr 0 Lr sorption heat pump. Energy Convers Manag 2015;90:175e83.
[23] Sayyaadi H, Ahmadi MH, Dehghani S. Optimal design of a solar-driven heat
! ! engine based on thermal and ecological criteria. J Energy Eng 2014.
L1 Rs Ma0 L2 Rs Ma0 [24] Ya I, Braslavsky, Kostylev AV, Stepanyuk DP. Optimization of starting process
a1 ¼
þ ; a2 ¼ þ of the frequency controlled induction motor. In: IEEE power electronics and
L21
L22 Tr ðL1 þ L2 Þ L21
L22 Tr ðL1 þ L2 Þ motion control conference. EPE-PEMC; 2008. p. 1050e3.
[25] Polyanin Andrei D, Manzhirov Alexander V. Handbook of mathematics for
engineers and scientist. Chapman & Hall/CRC; 2007.