Rotor Flux Oriented Induction Machine as a DC Power Generator E.

Levi

Rotor Flux Oriented Induction Machine as a DC Power Generator

E. Levi, Y.W.Liao
LIVERPOOL JOHN MOORES UNIVERSITY
School of Engineering, Byrom St
Liverpool L3 3AF, UK
Tel: +(44-151) 231 2257, Fax: +(44-151) 298 2624
e-mail: e.levi@livjm.ac.uk

Keywords
Renewable energy systems, Adjustable speed generation systems, Vector control, Flux model.

Abstract
The paper deals with a variable-speed induction generator, aimed at supplying an autonomous dc
power system with constant dc voltage. Reactive power requirements of the induction generator are
met by voltage source converter, operated in current controlled mode. Control of the compensator is
achieved using rotor flux oriented control principles. As operation in both base speed region and
field-weakening region is anticipated, the generator is represented with a suitable saturated machine
model. Estimation of instantaneous rotor flux space vector is performed using a recently introduced
estimator that fully accounts for main flux saturation. Description of the system is provided and an
extensive simulation study is performed. Self-excitation under no-load conditions, step application
and variation of a dc load, and variable speed operation in both base speed and field weakening
regions are examined by simulation. It is shown that the proposed control system is capable of
maintaining constant dc voltage under all operating conditions. Excellent tracking of rotor flux
reference is achieved, indicating that proper rotor flux orientation is maintained at all times.

Introduction
Stand-alone applications of variable-speed squirrel-cage induction generators require that reactive
power is supplied to the generator by means of a reactive power compensator. Among many possible
solutions [1], the most attractive one seems to be a PWM voltage source converter [2-10], due to its
inherent excellent dynamic characteristics. Control of the PWM voltage source converter (VSC) is
nowadays realised using vector control techniques [2-10]. Both stator flux oriented [3-6,8,10] and
rotor flux oriented control [2,7,9] may be applied, with flux space position calculation using either
indirect method or a variety of direct schemes.

A vector controlled variable-speed induction generator can be interfaced with the grid by means of the
second PWM converter [6,9], or it can be used to supply an autonomous power system composed of
either ac load [4,8,10] or dc load [3,5,7]. The PWM VSC is operated with either current control in
stationary reference frame [3,4-6] or in rotational reference frame [2,7,9]. Regardless of the method of
current control and regardless of the flux space vector along which orientation is performed,
information on instantaneous flux space vector spatial position is crucial for correct operation of the
system. Methods used so far include stator flux position calculation from measured stator voltages
and currents [3,5,6], rotor flux position estimation using indirect feed-forward approach [2] and rotor
flux position calculation from measured stator currents and rotor speed [9].

The paper deals with a variable-speed, rotor flux oriented induction generator, aimed at supplying
constant dc voltage to the dc side load. Hysteresis current control of the PWM VSC in stationary
reference frame is applied. Rotor flux estimation is performed from measured stator currents and rotor
speed, on the basis of a modified rotor flux estimator that fully accounts for main flux saturation [11].
The need for application of such a modified estimator stems from the fact that operation in both base

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.1
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

speed region and field weakening region is analysed. Performance of the scheme is illustrated with
simulation results. It is shown that dc voltage is kept constant at all speeds and under all loading
conditions. Excellent tracking of reference flux by the rotor flux estimate is demonstrated, indicating
full compensation of main flux saturation variation.

Description of the System

The outlay of the system is shown in Fig. 1, where load is connected at dc side of the converter. The
converter provides reactive power for the generator, while generator supplies active power to the load
RDC. A battery E is used for start-up purposes. Inverter firing signals are obtained on the basis of
closed-loop hysteresis current control algorithm in stationary reference frame. Control of the PWM
VSC static reactive power compensator is depicted in Fig. 2 (battery of Fig. 1 is omitted). Induction
generator phase currents are measured and used for both vector control algorithm and for hysteresis
current control. Twice the hysteresis band equals 0.1 A (3% of the rated peak stator current). Rotor
speed and dc voltage are measured as well and used in the vector control algorithm. Rated rotor flux is
an independent input into the vector control block. Rotor flux reference is constant and equal to rated
rotor flux in the base speed region, while in the field weakening it reduces inversely proportionally to
the speed. Outputs of the hysteresis current control blocks are shown as switching functions for each
of the three inverter legs of Fig. 1.

Vector control algorithm is illustrated in Fig. 3 and it consists of two branches that operate in parallel
and enable creation of stator phase current references. Current reference for d-axis is created by
closed loop rotor flux amplitude control. This control loop controls reactive power flow in the system.
The second branch creates q-axis current reference by closed loop control of the capacitor voltage vDC.
This control loop enables active power flow from the generator to the dc circuit and enables supply of
generated active power to the resistive load RDC. Output of the PI dc voltage controller represents
active power demand, which is divided with measured speed of rotation in order to create an
appropriate torque demand. Torque command is further divided with a signal obtained from the rotor
flux estimator, so that stator q-axis current command is created. Inversion of the sign of the q-axis
current reference accounts for the fact that active power flow is from ac to dc side of the converter.
Control system operates in rotor flux oriented reference frame and information about rotor flux space
vector instantaneous angular position is obtained from the block ‘Calculation of rotor flux and
position’, whose structure is discussed later.

As the available battery voltage for start-up of the generator is low (48 V), self-excitation has to be
initiated at a low speed [3,4]. Dc voltage reference is therefore during initial self-excitation taken as
linearly proportional to the speed of rotation, as illustrated in Fig. 4. Once when the generator
speed reaches rated synchronous value (1 p.u.), dc voltage reference is switched to a constant, speed-

I G
I G PWM VSI vDC R DC C

ω ω ia
Inverter firing ib Sa Sb Sc vDC
signals ic
Current control
R DC

C ω i a* ib * ic *

Vector control
rb E algorithm ψ rn

Current controlled
PWM VSI

Fig. 1: Induction generator with PWM VSC. Fig. 2: Structure of the vector control scheme.

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.2
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Calculation of rotor flux Rotor

amplitude and position speed
from measured signals Stator
currents
ψre φre
ω vDC* = k ω
ψrn ψr* − ids* ia*
PI 2
k at ω = 1 p.u.

ω jφre ib* vDC*

e
vDC* P* Te* iqs* ic* vDC* = 500 V = const.
PI −1 3
−
vDC ω (3/2)P(Lme/Lre)ψre

Fig. 3: Configuration of the block ‘Vector Fig. 4: Creation of the reference for dc voltage
control algorithm’ of Fig. 2. in Fig. 3.

independent value (500 V) and is further on kept unchanged under all operating conditions. Constant
k has such a value that dc voltage reference at 1 p.u. speed equals 500 V.

Rotor flux estimation is performed in rotor flux oriented reference frame, from measured stator
currents and rotor speed and the estimator fully accounts for main flux saturation. It is illustrated in
Fig. 5 (superscript e denotes estimated values, while asterisk stands for reference values). Detailed
derivation and description of this estimator is available in [11], where its application was addressed in
conjunction with a rotor flux oriented induction motor drive. Apart from calculation of rotor flux
amplitude and rotor flux spatial position for the vector control algorithm of Fig. 3, the estimator
additionally provides signal used to create stator q-axis current reference from torque reference in the
dc voltage control channel.

Simulation Procedure and Results

As operating regimes of the system include self-excitation process and field weakening region, the
induction generator is represented with a full saturated machine model of [12], with stator current and
rotor flux d-q axis components as state-space variables. The generator model is formed in the
reference frame rotating at constant 2π50 rad/s angular speed. A small 0.75 kW machine, whose data
and the magnetising curve approximation are given in Appendix, is used in simulations. Reference dc
voltage is set to 500 V and capacitor rating is 250 µF.

ia idse 1 ψre
3/2
ib and 1 + Tr e s
−jφre ωsle
ic e iqs e
1 φre
s
Lσr* ω
Lσr* _
Tr e 3P/2

Ψ2 Lre=f(Ψ2) (Rr*)−1
Lσr*
Lre
(3P/2) ψr (Lm / Lr )
e e e

Fig. 5: Configuration of the saturation adaptive rotor flux estimator.

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.3
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Battery voltage is normally small and is here taken as 48 V. Self-excitation process is therefore
initiated at 0.1 p.u. speed under no-load conditions. Turbine is not modelled and speed of rotation is
taken as an independent input. The generator runs at constant 0.1 p.u. speed (2π5 rad/s electrical)
from t = 0 s till t = 0.2 s. Linear increase in speed is initiated at t = 0.2 s and generator speed reaches 1
p.u. at t = 0.5 s. Dc voltage reference linearly increases during this time interval as well (Fig. 4) and is
switched to constant speed-independent 500 V value at t = 0.5 s.

At time instant t = 1 s resistive load of 500 Ω is connected at the dc side in a step-wise manner
(approximately 2/3 of the rated power for motoring). During time interval from t = 2 s to t = 2.1 s
speed is linearly decreased from 1 p.u. to 0.8 p.u. and is kept at 0.8 p.u. until t = 2.9 s. From t = 2.9 s
to t = 3 s speed is linearly changed from 0.8 p.u. to 1 p.u. Finally, load is altered from 500 Ω to 750 Ω
at t = 3.5 s. Figure 6 summarises simulation results for the described sequence of transients.

Dc voltage, fundamental component of stator phase voltage, estimated and actual rotor flux, estimated
and actual torque, actual and estimated magnetising inductance, and generator stator d-q axis current
components within the estimator of Fig. 5 are shown. As can be seen from Fig. 6, excellent matching
between actual and estimated values of both rotor flux (Fig. 6c) and electromagnetic torque (Fig. 6d)
is achieved, due to correct tracking of the actual magnetising inductance (Fig. 6e) provided by the
estimator. Torque response to load application and speed change is almost instantaneous and it does
not exhibit any oscillations, indicating that proper field orientation is maintained under all operating
conditions. Voltage at the dc side is at required 500 V value in all steady-states that follow after self-
excitation process. Change of generator speed by ± 20% hardly causes any dc voltage variation. Dc
voltage recovery after application of the load and after partial load removal is very quick. While peak
PWM stator phase voltage remains constant in all the steady-states after t = 0.5 s, fundamental
component (Fig. 6b) reduces when speed is reduced, as expected.

As dc side load is required to be supplied with constant dc voltage at all times, induction generator is
essentially operating with constant power demand for the given load resistance, regardless of the
speed of operation. From the point of view of generator’s utilisation, it appears to be optimal to
supply such a load by operating the machine in the field-weakening region (constant power region).
Simulation is therefore repeated once more, with the same sequence of events. The only difference is
that now during time intervals from t = 2 s to t = 2.1 s and from t = 2.9 s to t = 3 s speed at first
linearly increases from 1 p.u. to 1.2 p.u. and then returns from 1.2 p.u. to 1 p.u. Results are given in
Fig. 7, where the same graphs are displayed as in Fig. 6.

There is hardly any difference between dc voltages given in Figs. 6a and 7a. However, fundamental
component of the stator phase voltage differs: it remains unchanged during speed increase and
operation at 1.2 p.u. speed (Fig. 7b). This is consequence of operation in the field weakening region.
Actual and estimated values of both rotor flux (Fig. 7c) and electromagnetic torque (Fig. 7d) are again
in excellent agreement; this is enabled by correct adaptation of the estimator to the actual saturation
level in the machine, as indicated by the graph showing magnetising inductances (Fig. 7e). While
torque and hence stator q-axis current component had to increase in order to meet the same power
demand when speed was lowered from 1 p.u. to 0.8 p.u. in Fig. 6 (Figs. 6d and 6f), the situation is
now quite different. Constant power demand during speed increase from 1 p.u. to 1.2 p.u. is now met
with reduced torque value (Fig. 7d) while stator q-axis current component remains almost constant.

Figures 6 and 7 lead to the following conclusion. If induction generator is operated at all times at
speeds equal to or higher than 1 p.u., then power rating of the induction generator has to be essentially
the same as the maximum anticipated power demand at the dc side. However, if induction generator is
operated at speeds lower than 1 p.u., then generator rating must significantly exceed maximum power
demand at the dc side, in order to enable supply of this demand at the lowest allowed speed of
operation. This is a consequence of the need for increased electromagnetic torque (i.e., stator current)
when speed of operation is below 1 p.u.. It follows that the optimal case corresponds to operation of

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.4
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Capacitor d.c. voltage (V) Actual and estimated torque (Nm)

600 2

0
500

-2
Actual
400
-4

300 -6 Estimated
-8
200

-10

100
-12

0 -14
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (S) Time (s)

a) d)
Fundamental component of stator phase voltage (V) Actual and estimated Lm (Nm)
400 0.8

300 0.7

200 0.6

100 0.5 Actual

0 0.4
Estimated
-100 0.3

-200 0.2

-300 0.1

-400 0
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s) Time (s)

b) e)
Actual and estimated rotor flux (Wb) Estimator d-q axis current (A)
1
4
Estimated
0.9
3 ids
0.8
Actual 2

0.7 1

0.6 0

0.5 -1 iqs
0.4 -2

0.3 -3

0.2 -4

0.1 -5

0 -6
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s) Time (s)

c) f)
Fig. 6: Self-excitation, step application and alteration of load, and variable speed operation in the base
speed region: a) dc voltage, b) fundamental component of the stator phase voltage, c) estimated
and actual rotor flux, d) estimated and actual torque, e) magnetising inductance variation in the
generator and in the estimator, f) stator current d-q axis components in rotor flux oriented
reference frame.

the generator at all times at speeds equal to or higher than 1 p.u. (i.e., in the field-weakening region).

The need for utilisation of a saturation adaptive rotor flux estimator when the generator is operated in
the field-weakening region is verified by performing once more simulation identical to the one whose
results are given in Fig. 7. However, instead of the rotor flux estimator of Fig. 5, the corresponding

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.5
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Capacitor d.c. voltage (V) Actual and estimated torque (Nm)

600 2

0
500

-2 Actual
400
-4

300 -6 Estimated
-8
200

-10

100
-12

0 -14
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Tim e (s ) Tim e(s)

a) d)
Fundamental component of stator phase voltage (V) Actual and estimated Lm (H)
400 0.8

300 0.7

200 0.6

100 0.5
Actual
0 0.4
Estimated
-100 0.3

-200 0.2

-300 0.1

-400 0
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Tim e (s ) Tim e (s )

b) e)
Actual and estimated rotor flux (Wb) Estimator d-q axis current (A)
1 4

0.9 3 ids
0.8 Estimated 2

0.7 1
Actual
0.6 0

0.5 -1 iqs

0.4 -2

0.3 -3

0.2 -4

0.1 -5

0 -6
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4

Tim e (s) Tim e (s )

c) f)

Fig. 7: Self-excitation, step application and change of load, and variable speed operation in field
weakening region (traces as in Fig. 6).

constant parameter one is used, with magnetising inductance set to the constant rated value, given in
Appendix. Simulation results for this case are given in Fig. 8.

Dc voltage (Fig. 8a) is now significantly more disturbed during speed variation, compared to the case
illustrated in Fig. 7a. Fundamental component of the stator voltage (Fig. 8b) shows an increased value
during operation at 1.2 p.u. speed. This is a consequence of the discrepancy between actual and
estimated rotor flux (Fig. 8c) that takes place due to omission of main flux saturation compensation in

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.6
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Capacitor d.c. voltage (V) Actual and estimated torque (Nm)

600 2

0
500
Estimated
-2
400

-4
300 Actual
-6

200
-8

100 -10

0 -12
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s) Time (s)

a) d)
Fundamental component of stator phase voltage (v) Actual and estimator Lm (H)
400 0.8

300 0.7

200 0.6
Actual
100 0.5

0 0.4
Rated
-100 0.3

-200 0.2

-300 0.1

-400 0
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s) Time (s)

b) e)
Actual and estimated rotor flux (Wb) Estimator d-q axis currents (A)
1 4

0.9 3
Actual ids
0.8 2

0.7 1
Estimated
0.6
0
0.5
-1
iqs
0.4
-2
0.3
-3
0.2
-4
0.1
-5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 -6
Time (s) 0 0.5 1 1.5 2 2.5 3 3.5 4
Time (s)

c) f)

Fig. 8: Self-excitation, step application and change of load, and variable speed operation in field
weakening region with constant parameter rotor flux estimator (traces as in Fig. 7).

the estimator. Magnetising inductance in the estimator is constant and rated, while the one in the
machine significantly increases during operation in the field-weakening region (Fig. 8e). Reduction in
stator d-axis current in the field-weakening (Fig. 8f) is insufficient as the estimator does not recognise
change in the magnetising inductance. Actual and estimated values of the electromagnetic torque (Fig.
8d) are now different for operation at 1.2 p.u. Difference between results in Figs. 7 and 8 verifies the
need for application of a saturation adaptive rotor flux estimator.

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.7
Rotor Flux Oriented Induction Machine as a DC Power Generator E.Levi

Conclusion
Operation of a variable speed, rotor flux oriented induction generator as a source of constant dc
voltage is analysed. A saturation adaptive rotor flux estimator is employed in order to enable correct
operation of the scheme at speeds higher than rated. It is shown that the proposed control scheme is
capable of maintaining required dc voltage during speed variation in both base speed and field
weakening region and that dc voltage recovery after step load application and step load reduction is
very fast. Estimated and actual values of both rotor flux and torque are found to be in excellent
agreement for all operating conditions. The need for application of a saturation adaptive rotor flux
estimator is verified by investigating the behaviour of the same system in which a constant parameter
(rather than saturation adaptive) rotor flux estimator is used.

References
[1] D.W.NOVOTNY, D.J.GRITTER, G.H.STUDTMANN. Self-excitation in inverter driven induction
machines, IEEE Trans. on Power App. and Syst., vol. PAS-96, no. 4, 1977, pp. 1117-1125.
[2] C.B.JACOBINA, E.R.C. da SILVA, A.M.N.LIMA, R.L.A.RIBEIRO. Induction generator static systems
with a reduced number of components, Proc. IEEE Ind. Appl. Soc. Annu. Meet. IAS, San Diego, CA, 1996,
pp. 432-439.
[3] S.R.SILVA, R.O.C.LYRA. PWM converter for excitation of induction generators, Proc. 5th Eur. Conf. on
Power Elec. and Appl. EPE, IEE Conf. Pub. 377, Brighton, UK, pp. 174-178.
[4] R.O.C.LYRA, S.R.SILVA, P.C.CORTIZO. Direct and indirect flux control of an isolated induction
generator, Proc. IEEE Int. Conf. on Power Elec. and Drive Systems PEDS, Singapore, 1995, pp. 140-145.
[5] S.R.SILVA, R.O.C.LYRA. Pulse width modulation exciter for flux control of an isolated induction
generator, Proc. 6th Eur. Conf. on Power Elec. and Appl. EPE, Seville, Spain, 1995, pp. 3.233-3.238.
[6] G.JOHN, W.ERDMAN, R.HUDSON, C.S.FAN, S.MAHAJAN. Stator flux estimation from inverter
switching states for the field oriented control of induction generators, Proc. IEEE Ind. Appl. Soc. Annu.
Meet. IAS, Orlando, FL, 1995, pp. 182-188.
[7] C.COLLIEZ, A.TOUNZI, F.PIRIOU. Vector control of an autonomous induction generator connected to a
PWM rectifier, Proc. 7th Eur. Conf. on Power Elec. and Appl. EPE, Trondheim, Norway, 1997, pp. 2.711-
2.716.
[8] M.S.MIRANDA, R.O.C.LYRA, S.R.SILVA. Reactive static compensator for wind electric pumping system
using induction machines, Proc. IEEE Int. Conf. Electric Machines and Drives EMD, Milwaukee, WI,
1997, pp. MC2-8.1-MC2-8.3.
[9] M.G.SIMOES, B.K.BOSE, R.J.SPIEGEL. Design and performance evaluation of a fuzzy-logic-based
variable-speed wind generation system, IEEE Trans. on Ind. Appl., vol. 33, no. 4, 1997, pp. 956-965.
[10] M.S.MIRANDA, R.O.C.LYRA, S.R.SILVA. Wind-powered pumping system using induction machines
with pulse width modulation (PWM) excitation, Proc. 7th Eur. Conf. on Power Elec. and Appl. EPE,
Trondheim, Norway, 1997, pp. 2.637-2.641.
[11] E.LEVI, M.SOKOLA. A novel saturation adaptive rotor flux estimator for rotor flux oriented induction
machines, Proc. 7th Eur. Conf. on Power Elec. and Appl. EPE, Trondheim, Norway, 1997, pp. 1.518-1.523.
[12] E.LEVI. A unified approach to main flux saturation modelling in d-q axis models of induction machines,
IEEE Trans. on Energy Conversion, vol. 10, no. 3, 1995, pp. 455-461.

Appendix: Induction Machine Data

0.75 kW 380 V 2.1 A 50 Hz star 2P=4

R s = 10 Ω Rr = 6.3 Ω Lσs = 43 mH Lσr = 40 mH Lmn = 0.42 H

Im
Magnetizing curve approximation (rms values): Ψm = 0.86427 ⋅ 0.59976 ⋅ Im
1.1211
, Lm = Ψm I m
Dynamic inductance approximation ( L = dΨm dI m ):
Im 0.1211 1.1211
L = 0.86427 ⋅ 0.59976 ⋅ (1121
. ⋅ Im − 0.511226 ⋅ I m )

_______________________________________________________________________________________________________________
EPE’99 - Lausanne P.8