13

CHAPTER 2

DERIVATION OF STATE EQUATIONS AND PARAMETER DETERMINATION

OF AN IPM MACHINE

2.1 Derivation of Machine Equations

A model of a 3 phase PM machine is shown in Figure 2.1. Both the abc and the dq

axes are shown in the figure. The magnetizing current, due to the presence of the magnet, is

represented in Figure 2.1 by the current source if . The fictitious current source ( along with

the associated field winding Lmd) will be used in the derivation to follow to obtain an

expression for the flux which is equivalent to the flux created by the presence of the magnet

It is assumed in the following derivation that the rotor speed of the machine is

constant . Thus, no current flows in the damper windings contained within the rotor and

their presence may be ignored. It is also assumed that the machine is balanced.

The stator voltage equations of the IPM are [31]

d
va = λ a + ia r1
dt
d
vb = λ b + i b r 1
dt (2.1)
d
vc = λ c + i c r 1 ,
dt

where va , vb , vc are the a,b,c stator terminal voltages , λa , λb ,λc are the flux linkages in

the abc plane, ia , ib ,ic are the abc stator currents, and r1 is the stator resistance of each phase
 14

Figure 2.1. Model of a brushless PM synchronous machine

 15

of the stator winding. Thus, the assumption has been made that the resistances for each

phase of the stator are the same.

The flux linkages expressed in terms of the stator and terms of the current are

⎡λ a ⎤ ⎡ L aa L ab Lac ⎤ ⎡i a ⎤ ⎡ Lafd 0 0 ⎤ ⎡i f ⎤
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ λ b ⎥ = ⎢ Lba Lbb Lbc ⎥ ⎢ib ⎥ + ⎢ Lbfd 0 0 ⎥ ⎢ 0 ⎥ . (2.2)
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣⎢ λ c ⎦⎥ ⎣⎢ Lca Lcb Lcc ⎥⎦ ⎢⎣ i c ⎥⎦ ⎣ Lcfd 0 0 ⎥⎦ ⎢⎣ 0 ⎥⎦
⎢

The inductances given in Equation (2.2) will be kept in their symbolic form, transformed

into their equivalent dq axis counterparts, and then be defined in terms of the physical

construction of machine.

Due to the salient nature of the rotor, the inductances of Equation (2.2) are functions

of the position of the rotor and (assuming that the rotor is spinning) are functions of time.

This means that the inductance parameters are constantly changing - making the analysis of

the machine very difficult in its present form. A transformation commonly referred to as

Park’s transformation allows the equations describing the machine to be transformed into a

reference frame where the inductances are not functions of time. The reference frame

chosen to accomplish this goal is dependent upon the type of machine being looked at. For a

synchronous machine, this reference frame is that of the rotor. In other words, the stator

voltages, currents, and inductances will be projected onto the rotor side of the machine and,

thus, be rotating at the same speed with which the rotor is spinning. With this

accomplished, the inductances no longer vary with the position of the rotor.
 16

The transformation of a 3 phase balanced voltage, current, or flux linkage from the

hqdo = T( θ ) habcs , (2.3)

abc to the dq reference plane may be expressed as [32]

⎡ ⎤
⎢cos( θ ) 2π 2π ⎥
⎢ cos( θ - ) cos( θ + )⎥
⎢ 3 3 ⎥
2 ⎢ ⎥
T( θ ) = ⎢ sin( θ ) 2π 2π ⎥ ,
3⎢ sin( θ - ) sin( θ +
3 ⎥⎥
)
⎢ 3
⎢ ⎥
⎢ 1 1 1⎥
⎢⎣ 2 2 2 ⎥⎦

where

and θ denotes the reference frame chosen. For the rotor reference frame, θ=θr . Expressing

the q d and o terms individually and substituting θ=θr gives

2 2π 2π
hq = [ ha cos( θ r ) + hb cos( θ r - ) + hc cos( θ r + )]
3 3 3
2 2π 2π
hd = [ ha sin( θ r ) + hb sin( θ r - ) + hc sin( θ r + )] (2.4)
3 3 3
1
h o = [ h a + hb + h c ] ,
3

where, for balanced conditions, h0 = 0 and the o term will, from this point forward, be

ignored. So, using Equation (2.4), the flux linkages may be expressed in the dq plane as
 17

2 2π 2π
λ q = [ λ a cos( θ r ) + λ b cos( θ r - ) + λ c cos( θ r + )]
3 3 3
(2.5)
2 2π 2π
λ d = [ λ a sin( θ r ) + λ b sin( θ r - ) + λ c sin( θ r + )] .
3 3 3

d λq 2 dθ 2π dθr 2π dθr
= - [ λ a sin( θ r ) r + λ b sin( θ r - ) + λ c sin( θ r + ) ]
dt 3 dt 3 dt 3 dt (2.6)
2 d λa d λb 2π d λc 2π
+ [ cos( θ r ) + cos( θ r - )+ cos( θ r + )] .
3 dt dt 3 dt 3

Taking the derivative of λq in Equation (2.5) with respect to time gives

Comparing the first bracketed expression of Equation (2.6) with the expression for λd in

Equation (2.5) , gives

dθr dθ 2π dθr 2π dθr

- λd = [ λ a sin( θ r ) r + λ b sin( θ r - ) + λ c sin( θ r + ) ] . (2.7)
dt dt 3 dt 3 dt

Rearranging the voltage equation of (2.1) in terms of the rate of change of flux linkage

gives

d λa
= va - i a r 1
dt
d λb
= vb - i b r 1 (2.8)
dt
d λc
= vc - ic r 1 .
dt

Substituting Equation (2.8) into the second bracketed expression of (2.6) gives
 18
d λa d λb 2π d λc 2π
cos( θ r ) + cos( θ r - )+ cos( θ r + )=
dt dt 3 dt 3
2π 2π
[ v a cos( θ r ) + vb cos( θ r - ) + vc cos( θ r + )] -
3 3 (2.9)
2π 2π
[ i a r 1 cos( θ r ) + ib r 1 cos( θ r - ) + i c r 1 cos( θ r + )]
3 3
= v qs - i qs r 1

Substituting the results of Equations (2.7) and (2.9) into (2.6) gives

d λq d θr
= - λd + v qs - i qs r 1 . (2.10)
dt dt

Letting

d dθr
p= , and ω r = ,
dt dt

then

p λq = - λd ωr +vqs - iqs r1 . (2.11)

Therefore, the q-axis voltage can be written as

vqs = p λq + iqs r1 + λd ωr . (2.12)

Similarly, taking the derivative of the d-axis flux linkage of Equation (2.5) gives
 19
d λd 2 dθ 2π dθr 2π dθr
= [ λ a cos( θ r ) r + λ b cos( θ r - ) + λ b cos( θ r + ) ]
dt 3 dt 3 dt 3 dt
(2.13)
2 d λa d λb 2π d λc 2π
+ [ sin( θ r ) + sin( θ r - )+ sin( θ r + )] .
3 dt dt 3 dt 3
dθr dθ 2π dθr 2π dθr
λq = [ λ a cos( θ r ) r + λ b cos( θ r - ) + λ c cos( θ r + ) ] . (2.14)
dt dt 3 dt 3 dt

Comparing the first bracketed expression in Equation (2.13) with the equation for q axis flux

linkage in Equation (2.5) gives the result

Substituting the results of Equation (2.8) into the second bracketed expression of Equation

(2.13) gives

d λa d λb 2π d λc 2π
sin( θ r ) + sin( θ r - )+ sin( θ r + )=
dt dt 3 dt 3
2π 2π
[ v a sin( θ r ) + vb sin( θ r - ) + vc sin( θ r + )] -
3 3
2π 2π
[ i a r 1 sin( θ r ) + ib r 1 sin( θ r - ) + i c r 1 sin( θ r + )] (2.15)
3 3

= vds - i ds r 1 .

Substituting the results of Equations (2.14) and (2.15) into Equation (2.13) gives

d λd d θr
= λq + vds - i ds r 1 . (2.16)
dt dt
 20
After rearranging Equation (2.16) and substituting Equation (2.10) into it, the final

result for the d axis voltage is

vds= p λd +ids r1 - λq ωr . (2.17)

In order to obtain the individual self and mutual inductance terms ( given in Equation

(2.2 )), the parameters may be defined in terms of the physical construction of the machine ,

i.e. number of windings, physical dimensions, etc. , and then the equations obtained may be

transformed into the dq reference frame. Alternatively, as was done in this thesis, the flux

linkage equations may be transformed directly into the dq rotor reference frame ( θ equals

θr) and the q and d axis inductances may be defined directly in that frame of reference.

Thus,

⎡λ q ⎤ ⎡λ as⎤ ⎡Laa Lab Lac⎤ ⎡iqs⎤ ⎡Lafd Ladr Laqr⎤ ⎡ i f ⎤

⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢λ d ⎥ = T(θ ) ⎢λbs⎥ = T(θ ) ⎢ Lba Lbb Lbc⎥ (T(θ ) )-1 ⎢ids⎥ + T(θ ) ⎢ Lbfd Lbdr Lbqr⎥ ⎢idr⎥ , (2.18)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢⎣ λo⎥⎦ ⎣⎢λcs⎦⎥ ⎣⎢ Lca Lcb Lcc⎦⎥ ⎢ ⎥
⎣ io ⎦ ⎢
⎣ Lcfd Lcdr Lcqr⎥⎦ ⎢⎣iqr⎥⎦

where

⎡ L aa L ab Lac ⎤ ⎡ Lls + Lmq 0 0⎤

⎢ ⎥ ⎢ ⎥
T( θ ) ⎢ Lba Lbb Lbc ⎥ (T( θ ) ) = ⎢ 0⎥ ,
-1
0 Lls + Lmd
⎢ ⎥ ⎢ ⎥
⎣⎢ Lca Lcb Lcc ⎥⎦ ⎢⎣ 0 0 Lls ⎥⎦
 21

and

⎡ L afd 0 0 ⎤ ⎡ Lmd 0 0⎤
⎢ ⎥ ⎢ ⎥
T( θ ) ⎢ Lbfd 0 0 ⎥ = ⎢ 0 Lmdr 0⎥ .
⎢ ⎥ ⎢ ⎥
⎢⎣ Lcfd 0 0 ⎥⎦ ⎢⎣ 0 0 L mqr ⎥⎦

It is worthwhile to reiterate that the field current if is fictitious and has been used to

represent the flux linkage from the magnet source. From this point forth the term λe will be

used instead of the product term if * Lmd . Since the rotor speed is assumed constant, idr and

iqr are equal to zero.

The flux linkage equations in the dq axis can be expressed individually as

λ d = Ld i ds + λ e
(2.19)
λ q = Lq i qs ,

where

Ld = Lls + Lmd
L q = Lls + Lmq ,

and Lmd and Lmq are the d and q axis mutual inductances and L ls is the leakage in the

stator.
 22
The d and q axis mutual inductances are given as [12]

Lmq = C q Lm
(2.20)
L md = C d Lm ,

where Lm is the inductance of a machine with a uniform air gap and no magnets. This

inductance is determined from the flux linking with N1 Cw effective turns, and is given as

2
N 1 C w Di L - 8 H ,
L m = 1.273 µ o m1 ( ) 10 (2.21)
P g

where all the terms given in Equation (2.21) are in meter, kilogram, second (MKS) units

µo = permeability of free space

m1 = number of phases of the machine

Ni = number of series turns per phase

Cw = a winding factor which is a product of the distribution and pitch factors

Di = stator inner diameter

P = number of poles

L = core length

g = effective air gap length.

Cq and Cd are factors which account for the presence of the magnets and are, for an interior

permanent magnet, given as

 23
ρπ
C q = ρ - sin
π
8 π
( π 2 ) sin 2 ρ (2.22)
ρπ ρ 2
C d = ρ + sin - ,
π R
1+ g
Rm

where

ρ = the pole arc

Rg = reluctance of the air gap

Rm = reluctance of the magnet .

The open circuit magnet flux λe for an IPM machine is given as [12]

4.44 ( π Di L) !
λe = Ni Cw
-8
B f * 10 , (2.23)
2π P

where B!f is the amplitude of the fundamental flux density created by an individual magnet.

In summary, the voltage and flux equations needed to analyze a permanent magnet

under the stated assumptions are

v qs = p λ qs + i qs r s + λ ds ω
(2.24)
v ds = p λ ds + i ds r s - λ qs ω ,

where

λ ds = Lds i ds + λ e
λ qs = L qs i qs ,
 24
and the subscript s on the q and d axis stator terms has been added, rs has replaced r1 as

the symbol used to represent the stator resistance, and the subscript r has been dropped from

ω. Under steady-state, the derivatives of the state variables of Equation (2.24) are zero so, at

steady state the voltage equations may be written as

V qs = I qs r s + λ ds ω
(2.25)
V ds = I ds r s - λ qs ω .
A d-q axis schematic diagram representing the equations given in (2.25) is shown in Figure

2.2

Figure 2.2. Schematic diagram representing the steady state q and d axis voltage equations

of a PM machine
 25
2.2 Determination of Parameters of the IPM

The single most important determination to be made in the analysis of an electrical

machine is to find what the parameters of the machine are. Due to the effects of saturation at

heavy load and demagnetization of the magnet at light loads, the parameters of the machine

change significantly as the load presented to it changes. Thus, the parameters (namely the

mutual inductances in the d and q axes and the flux linkage created by the magnet) must be

functions of the operating conditions.

Generally, when effects due to saturation are included in an analysis, the parameters

are made functions of either the total mutual flux λmm or the total stator current Is. The

decision was made to make the parameters functions of stator peak current because, when

the parameters were plotted as a function of flux, there were regions in which two possible

values of a particular parameter could be obtained for a single value of flux (which could be

problematic when solving equations).

The parameters are found by using the equations given in Equation (2.25) and

repeated below for the sake of continuity.

V qs = r s I qs + ω Lds I ds + ω λ e
(2.26)
V ds = r s I ds - ω Lqs I qs .
 26

Figure 2.3. Schematic diagram of dc test used to determine stator resistance

The stator resistive value rs was found by applying a dc voltage across two terminals

of the stator and measuring both the voltage and the current which flowed through the

terminals ( see Figure 2.3). The stator resistance for a single phase is given as

V dc
rs = . (2.27)
2 I dc

The voltages and currents in Equation (2.26) were found by varying a three phase

balanced resistive load from a high value to a low value and recording the terminal voltage

and the current output from the generator. The measurements made were the line to line

voltages and the phase currents. In order to convert the voltages and currents into their dq

components, the torque angle δ was needed. The power factor of the machine is also needed

in parameter determination, but, since the generator was feeding a resistive load, the current

out of the generator was in phase with the voltage at the terminals, so the power factor was
 27
unity. The torque angle was found by measuring the difference in angle of the voltages of

a search coil located across phase “a” of the stator and the terminal voltage appearing at the

stator terminal of phase “a.” This method was not an ideal way to measure the torque angle

because the oscilloscope used to measure the angle between the two voltages gave a varying

readout even though the load and speed of the generator were constant. An average of the

numbers was taken and used as the torque angle. It would have been much easier (and

probably more accurate) to have a commercially available torque angle measuring device;

however, no such device was available. Nevertheless, the strong corroboration between

measured and predicted results suggests that the method used was an acceptable means of

obtaining the torque angle.

Once the stator voltages, currents, and torque angle are known, the dq voltages and

V qs = V s Cos( δ )
V ds = - V s Sin( δ )
(2.28)
I qs = I s Cos( γ )
I ds = - I s Sin( γ ) ,

currents can be found by the following relations:

where Vs is the peak line to neutral voltage, Is is the peak stator current, and γ is the sum of

the torque angle δ and the power factor angle θ. Since the power factor is unity (since the

generator is feeding a purely resistive load), then γ is equal to δ.

With the dq voltages and currents and the stator resistance known, the inductance

in the q axis can easily be found and is given as

 28

V ds - r s I ds
L qs = - . (2.29)
ω I qs

The inductance in the d axis and the magnet flux linkage are not as easy to find as the q axis

inductance. The two terms are contained in the same equation and are, in a sense, coupled

together.

One method of finding the magnet flux involves running a no load test on the PM

machine for a range of frequencies and measuring the terminal voltage of the machine and

the voltage across the terminals of the search coil. An empirical relationship between the

search coil voltage and the magnet voltage (and thus the magnet flux linkage) can be

developed since, at a no load condition, the terminal voltage of the machine is equal to the

magnet voltage. Figure 2.4 shows a plot of the rms voltage of the magnet vs. the air gap

voltage for both the series connection (high voltage) and the parallel connection (low

voltage) of the stator winding of the PM machine.

The low voltage connection was not used in any of the experiments reported in this

thesis (except for the one just described). The main reason for this is that, since the machine

is being operated in generator mode, one would normally like a high terminal voltage and

the low voltage connection is, as one would expect, one half of the terminal voltage of the

high voltage connection for any particular operating frequency.

This brings up one other point of interest, which is that one would ideally like to

have a generator having a magnet voltage greater than 120 volts rms when operating at 60
 29

100

90 + E o h v = . 1 6 8 3 * V a g -. 1 1 2 8

80 o Eo l v =. 0 8 3 9 * Va g +. 1 8 3 5

70

60
Eo (V)

50

40

30

20

10

0
100 200 300 400 500 600
A i r g a p v o l t a g e ( mV )

Figure 2.4. Measured line to neutral rms terminal generator voltage vs air gap voltage for no

load condition for machine connected in high and low voltage stator connections
 30
Hertz since most loads were designed to operate at or near that particular voltage.

Although not shown explicitly on Figure 2.4, the machine is operating at 60 Hertz when the

air gap voltage is approximately 520 mv. This operating point corresponds to a magnet

voltage of about 88 volts line to neutral rms which is certainly low if one wanted to drive, for

example an induction motor with it. As will be seen in subsequent chapters, shunt capacitors

placed at the terminals of the PM machine boosts the terminal voltage, but a better method

would be to design a PM machine having a magnet voltage of around 130 volts when

operating at 60 Hertz. Incidentally, the lower voltage of this IPM machine is not an

indication of poor design; rather, it is an indication that it was designed to be used as a

motor.

The empirical relationship between the air gap voltage (Vag) and the magnet voltage on a per

phase rms basis (Eo ) was found to be

E o = 0.1683 * V ag - 0.1128 . (2.30)

With this relationship established, the magnet voltage (and thus the magnet flux) could be

approximated under load conditions by measuring the air gap voltage at each operating

condition. While this method is not entirely accurate since, under load conditions, the

voltage across the search coil is also affected by the current flowing in the mutual

inductances of the d and q windings of the stator, the approximation seems reasonable and,

absent the use of finite element analysis, there is little other option but to use this method if

the d axis inductance and the magnet flux terms are to both remain as functions of the

operating conditions.
 31
After the magnet flux term has been determined (by λ e = E o / ω), then Lds can be

found by

V qs - r s I qs - ω λ e
L ds = . (2.31)
ω I ds

The plots of Lqs, Lds, and λ e are given in Figures 2.5-2.8. The empirical relationships

of the parameters as a function of peak stator current Is are given as

1
ln( ) = - .0013 Is4 + .0293 I3s - .2303 Is2 + .8684 Is + .790
Lqs
1
ln( ) = - .0011 Is4 + .0251 I3s - .210 Is2 + .9096 Is + 1.505 (2.32)
Lds
λ e = .0002 I3s - .0041 Is2 + .0208 Is + .1863 .

In a conventional synchronous machine with a field winding, the d axis inductance is

larger than the q axis inductance; however, comparing the magnitudes of Lqs and Lds in

Figures 2.5 and 2.6, it can be seen that Lqs is larger. This phenomenon, called inverse

saliency, is caused by the magnet depth appearing as basically an air gap in the d-axis. As

the machine is loaded, it can also be seen that the magnet voltage Eo first increases and then

decreases. The initial increase in magnet voltage is due to the fact that, at very light loads,

the bridge becomes highly saturated and much of the magnet flux flows through it and does

not contribute towards a useful air gap voltage.

It was mentioned earlier in this section that the reason the parameters were made

functions of the peak stator current and not the total flux linkage was that the when the
 32

0.3

0.25 + E x p e r i me n t
- Fi t

0.2
Lqs [H]

0.15

0.1

0.05
0 1 2 3 4 5 6 7 8
Is p e a k [ A ]

Figure 2.5. Measured values of q axis inductance vs peak stator current

 33

0.12
+ E xp
0.1
- Fi t

0.08
Lds [H]

0.06

0.04

0.02

0
0 1 2 3 4 5 6 7 8
Is p e a k [ A ]

Figure 2.6. Measured d axis inductance vs peak stator current

 34

0.22

0.21

0.2
Ma g ne ti c F lux [Wb ]

+ E xp
0.19 - Fi t
0.18

0.17

0.16

0.15

0.14

0.13

0.12
0 1 2 3 4 5 6 7 8
Is p e a k [ A ]

Figure 2.7. Measured magnetic flux vs peak stator current

 35

0.3

0.25

0.2
Lqs (H)

0.15

0.1

0.05
0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32
M u t u a l F l u x [ Wb ]
Figure 2.8. Measured q axis inductance vs peak mutual flux
 36
parameters were plotted as a function of the total mutual flux linkage λmm, there were

regions in which two possible solutions exist. An example of this is shown in Figure 2.8

where the q axis inductance is plotted vs the mutual flux linkage. It can be seen from the

figure why determining Lqs from a given value of λmm would be difficult. For example, if

λmm was given as 0.28 Wb, then Lqs could either be 0.1 or 0.15 H. The same problem was

present for the parameters Lds and λe as well.