Optimal control of interior permanent magnet synchronous integrated

starter-generator
12

L. C H EDOT
G. F RIEDRICH2
1
2
Valeo Electrical System
Universite de Technologie de Compi`egne
2, rue A. Boulle / BP 150
Laboratoire dElectromecanique / BP20529
94017 Creteil Cedex / France
60205 Compi`egne Cedex / France
Email: laurent.chedot@utc.fr
Email: guy.friedrich@utc.fr
Tel: 33 (0)3 44 23 45 15
URL: www.utc.fr/lec

Keywords
Permanent magnet motors, automotive applications, drives for HEV, adjustable speed drives, flux model.

Abstract
This paper deals with the optimal control of interior permanent magnet synchronous machine (IPM) in
the integrated starter-generator (ISG) application. IPM designed for flux-weakening operations are able
to realise high performances. Nevertheless, the ISG environment imposes lots of constraints which must
be taken into account in an appropriate control. After a presentation of machine design, ISG constraints
and control structure, models of the IPM and its environment are detailed including electromechanical
calculations. An optimal control (total losses minimisation) based on a numerical, non linear constrained
optimisation routine is described. Efficiencies in torque vs. speed plan point out the importance of taking
into account voltage and power limitations against the operating range. The presented simulation results
shows that Interior Permanent Magnet machine (IPM) is a good solution for ISG applications.

1 Introduction
The study of ISG application leads to make comparisons between different machines structures: induction machine, wound rotor synchronous machine, reluctant and permanent magnet machine [16].
All these machines must respect very strong rules and specifications (low size, high torque, speed and
efficiency). In this context, IPM structure owns lots of advantages: high specific power, brushless, no
losses in the rotor. IPM particularities, associated to ISG constraints (wide speed range, battery supply
and highly variable temperature) impose a precise control.
After a presentation of machine design, ISG constraints and control structure, models of the IPM and
its environment will be detailed including electromechanical calculations. Then, optimisation procedure
will be established.

2 Machine and control structure

2.1

Design

Figure 1 shows a cross-section of a classical IPM adapted to flux-weakening operation [7, 8].
This structure cumulates the characteristics of permanent-magnet and reluctant machines [9]: torque
is a combination of hybrid and reluctant torque; the induce voltage, due to the presence of permanentmagnet excitation, is constant and must be reduced by flux-weakening at high speed. Electrical and
mechanical behaviour will be detailed in section 3.

Figure 1: IPM cross-section [7]

2.2

Constraints due to starter-generator application

Starter-generator, as others automotive applications, is very constrained:

low size;
high torque at low speed with minimum power taken on the battery (140 N m at 600 Arms , 8 kW );
operating points at high speed ( 6000 rpm);
power and voltage limited by battery: 8kW , 21 to 36 V in Motor mode (starter or boost) and 42
to 50V in Generator mode (power depending on battery technology);
limited battery energy storage;
current limited by inverter or thermic conditions (150 600 Arms );
high temperature variation (25o C 180o C).
These constraints create specific behaviours (high magnetic saturation) and limitations (current, voltage, power, energy). Moreover, terminal voltage, equal to the battery voltage, varies with the state of
charge and the consumed power [10].

2.3

Control scheme

This machine is used as starter and generator. Its control is unified by using of unique torque control,
positive torque for motor operations and negative torque for generator. Figure 2 shows this control
scheme.

Battery

id*
C*

Optimal
control
laws

IPM
PWM

Inverter

iq*

ia

ib

d/dt

Figure 2: Optimal torque control

The optimal control laws, including flux-weakening [11, 12], can be explained thanks to circle diagram and the three control modes introduced by M ORIMOTO and all. in [13]. In order to use simple
analytical expression, a lot of hypotheses must be done: no magnetic saturation, constant terminal voltage, no temperature variation.
We show that, in the starter-generator application, these hypotheses can not be maintained. In
these conditions, to realise a precise control including high efficiency, it become necessary to take
into account all the non-linearities (machine and application) in a specific control.
One way is to compute, by numerical calculations, M ORIMOTOs ideal trajectories with precise
models of the machine.

3 Models
3.1

Machine

The IPM is modelled by classical Parks equations [14] (d-q reference frame) except for flux and iron
power losses.
3.1.1 Saturation
Because of magnetic saturation, flux can not be expressed as functions of inductances. In each axe, the
inductance saturates with the current (classical saturation) and moreover each current has an action on
the others inductances (cross saturation).
Nevertheless, in the Parks equation, the relations linking flux with voltage and torque dont need hypotheses on magnetic saturation (Cf. equations 10, 13 and 14). These relations stay true in high magnetic
saturation conditions.
Each flux d and q can be a non-linear function of the currents id and iq .
d = fd (id , iq )

(1)

q = fq (id , iq )

(2)

fd and fq are calculated by interpolation of measures tables (Cf. figure 3) realised with the finite
element (FE) software FLUX2D [15].

Figure 3: Flux table

For different operating points (id , iq ), and in presence of the permanent magnets, the 3 phase flux
(a , b , c ) are evaluated (internal function) and so, direct and quadrature flux are deduced:

 
id
iq

P ark 1

ia
ib
ic

which give after resolution:

a
b
c

P ark

 
d

3.1.2 Iron losses

Iron losses evaluation follows the same procedure, FE measurements (FLUX2D) for different speed and
different operating points give tables of data which are interpolated
Piron = f (id , iq , s )

(3)

After the prototype construction, iron losses will be measured and the characteristics will be used
for the optimisation.

3.2

Inverter

Inverter behaviour is considered as ideal. Its global efficiency and load voltage drop can be added easily.

3.3

Battery

Battery is modelled by a voltage source in series with an internal

resistance as shown on figure 4. Maximum power supply is equal to:
Pbmax

E2
= b
4Rb

Rb
Ub

(4)

In generator mode, terminal voltage is regulated at a constant value

(around). Battery is so modelled by a simple voltage source Vch .

Eb

Ib

Figure 4: Battery electrical model

3.4

Electromechanical equations

For a given operating point (id , iq and s ), all electrical and mechanical data are performed:
current (RMS),
Irms =

i2d + i2q
3

(5)

flux (table),
d = d (id , iq )

(6)

q = q (id , iq )

(7)

iron (table) and total losses,

Piron = Piron (id , iq , s )
Plosses = Piron +

2
3Rs Irms

(8)
(9)

electromagnetic, losses and mechanical torque,

Tem = p [d iq q id ]
Piron
Tiron =

Tm = Tem Tiron

(10)
(11)
(12)

voltage,
vd = Rs id s q

vq = Rs iq + s d
s
vd2 + vq2
Vrms =
3

(13)
(14)
(15)

electrical and mechanical power,

Pe = vd id + vq iq
Pm = Tm
efficiency,
=
battery voltage,
Ub =

(17)

sign(Tm )

(18)

q
Eb2 4Rb Pb

(19)

Pm
Pe

Eb +

(16)

where Pb is the power given by the battery. It is equal to the electrical power divided by the
inverter efficiency. If voltage are considered as sinusoidal, across voltage supply is equal to:
Ub
Vsup =
2 2

(20)

The maximum injectable current is equal to the maximum inverter current in starter mode and is
limited by the maximum current density in the IPM in generator mode:

Iinverter max (starter)
Ilim =
(21)
Js max
(generator)

4 Optimisation procedure
4.1

Principle

Controlling the IPM is equivalent to injecting the currents id , iq which minimise the total losses with
respect to different constraints (torque, current, voltage and power).
(T , ),

(id , iq ) \ min

id ,iq

with

Plosses

Tm = T
Vrms Vdisp
Irms Ilim

Pe Pbmax

The M ATLAB optimisation toolbox [16] provides a non-linear constrained optimisation routine. It
minimise an objective function f and try to maintain constrained functions g negative:
x

min
f (x )

with
gi (x ) < 0,

4.2

i = 1..Nconstraints

Objective function

The objective function is here the total losses (Cf. equation 9):
f = Plosses

4.3

(22)

Constraints functions

The constraints functions are:

Mechanical torque (Cf. equation 12) is equal to the order torque:
gt = |Tm T | |T |

(23)

 is a percentage (0 <  < 1) which defines precision.

Current (Cf. equation 5) is less than the limit (Cf. equation 21):
gi = Irms Ilim

(24)

across voltage (Cf. equation 15) is less than the available voltage (Cf. equation 20)
gv = Vrms Vsup

(25)

In starter mode, electrical power (Cf. equation 16) is limited by the battery maximum power
(Cf. equation 4):
gp = Pe Pbmax
(26)

4.4

Algorithm

1. The operating range is established. Speed is due to the application: from 0 to the application
maximum speed (5000 rpm). Maximum and minimum torques are calculated at standstill by a
first constrained optimisation.
2. For each couple (T , ), (id , iq are calculated by optimisation as seen before.

5 Application
As an application, the currents are calculated and used to evaluate the IPM performances and operating
range.

5.1

Environmental data

The previous calculation procedure is done with the following data and limits:
temperature is close to the ambient temperature (T = 25o C), stator resistance is so equal to:
Rs = 6.1 m;
in starter mode, current is limited by inverter: Ilim = 600 Arms ; and voltage by battery: Eb =
36 V, Rb = 40 m;
in generator mode, current is limited by current density in IPM: Js 10 A/mm2 Ilim <
190 Arms ; and voltage by the regulation: Vsup = 2Vch2 , Vch = 42 V .

5.2

Results

We can see on figure 5 efficiencies in the torque vs. speed plan.

Efficiencies=f(W,T*)
140
Maximum battery power
Motor base point power
Generator base point power

120

0.8

100

0.7

80

0.6

T* (Nm)

60

0.5

40
0.4
20
0.3
0
0.2
20
0.1

40

60

500

1000

1500

2000

2500
3000
W (RPM)

3500

4000

4500

Figure 5: Efficiencies in torque vs. speed plan

5000

Positive torques represent the starter mode, and so negative torques, the generator mode. We can see
lots of differences between this two kind of operating mode:
maximum torque is limited by the highest injectable current and the maximum electrical power.
In generator, thermic conditions in steady-state limit current density;
base speed and operating limits are directly affected by the voltage limitation. Even with our
simple model of battery, the influence of the terminal voltage decreasing is very important in
terms of operating range;
in generator mode, some operating points are not accessible. For low torque or low speed, the
mechanical power is not sufficient to compensate the losses (copper and iron).

6 Conclusion
This paper has examined the principle of the optimal control of interior permanent magnet synchronous
machine in the starter-generator application. It has been shown that IPM particularities (permanent magnet, reluctant torque) require a precise control. The strong constraints of the ISG application, particularly
the magnetic saturation and the voltage supply, have highlighted the necessity of taking into account all
these non-linearities in an optimal control. This control was established by classical optimal calculations. The presented simulation results show that Interior Permanent Magnet machine is a good solution
for ISG applications.

References
[1] B.J. C HALMERS, L. M USABA, and D.F. G OSDEN. Variable-frequency synchronous motor drives for electrical vehicles. IEEE Trans. Ind. Appl., 32(4):896903, Jul./Aug. 1996.
[2] E.C. L OVELACE, T.M.. JAHNS, J.L. K IRTLEY, and J.H. L ANG. An interior PM starter-alternator for automotive applications. In Int. Conf. Electrical machines, pages 18021808, Istanbul, Sep. 1998.
[3] J.R. H ADJI -M INAGLOU and G. H ENNEBERGER. Comparison of different motor types for electric vehicle
application. EPE Journal, 8(3-4):4655, Sep. 1999.
[4] G. F RIEDRICH, L. C H E DOT, and J.M. B IEDINGER. Comparison of two optimal machine design for integrated starter-generator applications. In Int. Conf. Electrical machines, Aug. 2002.
[5] C. P LASSE, M. C HEMIN, G. L ACAMOIRE, and E. VON W ESTERHOLT. Lalterno-demarreur, du stop & go
au groupe motopropulseur hybride. In Congr`es de la Societe des Ingenieurs Automobile, Versaille, Nov.
2001.
[6] C. P LASSE, A. A KEMAKOU, P. A RMIROLI, and D. S EBILLE. Lalterno-demarreur, du stop & go au groupe
motopropulseur mild hybride. In PropElec, Aix-en-Provence, Mar. 2003.
[7] N. B IANCHI, S. B OLOGNANI, and M. Z IGLIOTTO. High-performance PM synchronous motor drive for an
electrical scooter. IEEE Trans. Ind. Appl., 37(5):13481355, Sep./Oct. 2001.
[8] C.C. H WANG and Y.H. C HO. Effects of leakage flux on magnetic fields of interior permanent magnet
synchronous motors. IEEE Trans. Mag., 37(4):30213024, Jul. 2001.
[9] T.J.E. M ILLER. Brushless Permanent Magnet and Reluctant motor drive. Oxford university press, 1989.
[10] E. K UHN. Mod`ele de batterie NiMH pour vehicule hybride parall`ele : validation en grands signaux. In
JCGE, Nantes, 2003.
[11] W.L. S OONG and T.J.E. M ILLER. Field-weakening performance of brushless synchronous AC motor drives.
IEE Proc. - Elec. Power Appl., 141(6):331340, 1994.
[12] B. M ULTON, J. L UCIDARNE, and L. P R E VOND. Analyse des possibilites de fonctionnement en regime de
desexcitation des moteurs a` aimants permanents. Journal Physique III, pages 623640, 1995.

[13] S. M ORIMOTO, Y. TAKEDA, T. H IRASA, and K. TANIGUCHI. Expansion of operating limits for permanent
magnet motor by current vector control considering inverter capacity. IEEE Trans. Ind. Appl., 26:866871,
Sep./Oct. 1990.
[14] G. S E GUIER and F. N OTELET. Electrotechnique industrielle. Technique et documentation, 1994.
[15] Cedrat. Analyse des dispositifs e lectriques, magnetiques et thermiques par la methode des e lements finis.
Notice dutilisation generale, Oct. 1996.
[16] The MathWorks Inc. Optimization toolbox users guide for use with Matlab, 2000.