B- H loop and demagnetization characteristics The starting-point for understanding magnet
3. Induction machines
3.8 T RANSIENTS AND EIGENVALUE ANALYSIS
PC-IMD can calculate starting transients. To model the capability of soft-starters with point-on-wave
switching, the AC supply voltage can be switched in two stages, so that the DC offset can be suppressed. This practically eliminates the oscillatory torque which occurs when the three lines close simultaneously, and considerably lightens the duty on the shaft coupling, [Wood, 1965].
PC-IMD can also calculate the eigenvalues of the motor operating from a sinusoidal voltage source at
zero slip over a range of frequencies from the normal line frequency down to zero. The resulting root- locus diagram is indicative of the relative stability of the motor when fed from a variable-frequency inverter without shaft position or velocity feedback.
The theory of these dynamic calculations is presented here.
Reference Frame Transformations
PC-IMD’s dynamic calculations use two sets of d,q axes:
1. synchronously rotating (Kron), for eigenvalue analysis; and
2. d,q axes fixed on the rotor (Park) for transient analysis (starting calculations).
The conventions are the same as in Fitzgerald and Kingsley [1961], with only one or two minor changes of notation: see Fig. 3.15.4 Although this approach is rather old-fashioned it is practical and clear. PC- IMD does not include zero-sequence voltages or currents, so the theory works only for three-wire
connections and the transformation matrices are not square.
The transformation from (a,b,c) variables (i.e., direct phase variables) to dq axes is eqn. (3.29), where 2 is the angle between the d-axis and the axis of phase a in elec. rad. The same transformation (and its inverse, eqn. (3.30)) is used for voltages, currents, and flux-linkages.
vd ' 2
3 [vacos 2 % vbcos (2 ! 2B/3) % vccos (2 % 2B/3)];
vq ' ! 2
3 [vasin 2 % vbsin (2 ! 2B/3) % vcsin (2 % 2B/3)].
vaia % vbib % vcic ' 3 2[vdid % vqiq] (3.31) Ra ' Laaia % Labib % Lacic % LaAiA % LaBiB % LaCiC Rb ' Lbaia % Lbbib % Lbcic % LbAiA % LbBiB % LbCiC Rc ' Lcaic % Lcbib % Lccic % LcAiA % LcBiB % LcCiC (3.32) LaA(2) ' LaA cos 2r LaB(2) ' LaA cos (2r % 2B/3) LaC(2) ' LaA cos (2r ! 2B/3) (3.33) Rq ' L11iq % L12iQ. (3.35) Rd ' 2
3[Ra cos 2 % Rb cos (2!2B/3) % Rc cos (2%2B/3)] ' L11id % L12iD (3.34)
RD ' L22iD % L12id;
RQ ' L22iQ % L12iq. (3.36)
Inverse:
At this stage it is not specified whether the dq axes are fixed to the rotor or to the stator or to the flux, or what. All that is necessary to specialize the transformation to one of the classic frames of reference is to constrain 2 appropriately.
The total electrical instantaneous power at the terminals is
Flux-linkages and inductances
There are three phases a, b, c on the stator and three on the rotor, A,B,C. For the stator,
The self-inductances are constant and symmetrical: Laa
= L
bb= L
cc;
LAA= L
BB= L
CC. The stator-stator
mutual inductances are constant: Lab
= L
bc= L
ca= L
ba= L
cb= L
ac. The rotor-rotor mutual inductances
are constant: LAB
= L
BC= L
CA= L
BA= L
CB= L
AC.
The stator-rotor mutual inductances vary sinusoidally with rotor position 2r
: for example,
A similar set of equations applies for LbA(2), LbB(2), LbC(2),LcA(2), LcB(2), and LcC(2), with appropriate angles taken from Fig. 3.15. Also an equation similar to eqn. (3.32) is written for the rotor flux-linkages RA
,R
B, and R
C.
The three stator phases will be replaced by two fictitious, orthogonal coils d and q fixed to (and rotating with) the d- and q- axes respectively. The flux-linkage of the d-coil is obtained by applying the voltage transformation to Ra
, R
band R
c:
where L11 = Laa ! Lab and L12 = 3/2 LaA and much algebra has been omitted. Similarly in the q-axis,
For the rotor D and Q coils, the transformation is the same but with 2r
instead of 2; thus
where L22 = LAA ! LAB, and again L12 = 3/2 LaA.
va ' vd cos 2 ! vq sin 2;
vb ' vd cos (2 ! 2B/3) ! vq sin (2 ! 2B/3);
vc ' vd cos (2 % 2B/3) ! vq sin (2 % 2B/3).
va ' Raia % pRa; vb ' Raib % pRb; vc ' Raic % pRc (3.37)
vd ' Rdid % pRd ! p2.Rq;
vq ' Rqiq % pRq % p2.Rd. (3.38)
Fig. 3.16 d-axis aligned with phase a
Physical attributes of the fictitious d and q coils
Consider the instant when the d-axis is aligned with the axis of phase a, and ia
= I, i
b= i
c=!½I in the 3-phase winding,
Fig. 3.16. According to the reference frame transformation
id
= I. If the flux-linkage in phase a is Q, the flux-linkages of
phases b and c must be !½Q each, so that by the reference eframe transformation Rd
=
2/3 × Q × [1 !½(!½)!½(!½)] =Q. The flux-linkage per ampere in the d-coil, is Rd
/
id= Q/I.
The d-coil evidently has the same number of turns as the a- coil. If we consider I to be a DC current then the voltage drop in phase a is va
= R
aIand the voltage-drops in phases
b and c are vb
= v
c= !R
aI/2. According to the reference frame transformation, v
d= 2/3 × R
aI× [1
!½(!½)!½(!½)] = RaI
. The resistance of the d-coil is evidently v
d/
id= R
a. This is consistent with
having the same number of turns, provided that the total cross-section of copper in the conductors of the d-coil is the same as that of the a-coil.
Consider the stator flux-linkage produced by the stator currents alone at the position shown in Fig. 3.16, 2r
= 0. R
ais made up of L
aaIdue to self-flux-linkage, plus (!½I) × L
abmutually coupled from each of
phases b and c, for a total of (Laa
!
Lab)
I= L
11I. The flux-linkage of phase b is L
aa× (!½I) due to self
flux-linkage, with Lab
× (!½I) mutually coupled from phase c, and L
abImutually coupled from phase
a, for a total of !½L11I
. The flux-linkage of phase c is the same. So R
d = 2/3 × L11I ×[1+(!½)×(!½)+(!½)×(!½)] = L11I
. According to the transformation equation the d-coil current is I, so
the apparent inductance is L11. The d-coil represents all three stator phase coils along the d-axis,
including their mutual coupling.
Now consider the stator flux-linkage produced by the rotor currents alone, with current I flowing in coil
A and !½I in coils B and C. The flux-linkage of stator coil a is Ra
= L
aAIfrom coil A and !½L
aA× (!½I)from each of coils B and C, for a total of LaAI
× [1 + (!½) × (!½) + (!½) × (!½)] = 3/2 × L
aAI. The flux- linkage of stator coil b is Rbcomprising !½L
aA× I from coil A, L
aA× (!½I) from coil B, and !½L
aA×
(!½I) from coil C for a total of !¾LaAI. The flux-linkage of coil c is the same, so that according to the
reference frame transformation Rd = 2/3 × [3/2 + (!½) × (!¾) + (!½) × (!¾)] × LaAI = 3/2 × LaAI
.
According to the transformation the current iD
is I × 2/3 × [1 + (!½) × (!½)+(!½) × (!½)] = I, so that
the apparent mutual inductance is 3/2 × LaA
.
Voltage equations
For the stator
where p is the operator d/dt, and we have assumed that Ra
= R
b= R
c. For the stator d- and q-coils the
voltage equations are derived by incorporating eqns. (3.37) into the reference-frame transformation equations for vd
and v
q. When p operates on the trigonometric functions of 2, we get the speed voltages.
Missing out a lot of algebra,
In synchronously rotating axes p2 = T = 2Bf. In d,q axes fixed to the rotor, p2 = Tr
,
i.e., the rotor angular velocity in elec rad/s. Note that Rd= R
q= R
a.
A similar procedure applies to the D- and Q-coils on the rotor, except that they are themselves rotating at angular velocity Tr
, the physical angular velocity of the rotor in elec rad/s. The result is
vD ' 0 ' RDiD % pRD ! p2s.RQ vQ ' 0 ' RQiQ % pRQ % p2s.RD. (3.39) Te ' 3 2 P 2 [Rdiq ! Rqid] [Nm]. (3.40) Te ' J dTr dt × 2 P % DTm (3.41) pRd ' vd ! Rdid % TRq pRq ' vq ! Rqiq ! TRd pRD ' ! RDiD % (T ! Tr)RQ pRQ ' ! RQiQ ! (T ! Tr)RD pTr ' P 2J [Te ! DTm] (3.42) va ' vpk cos (Tt % *); vb ' vpk cos (Tt % * ! 2B/3); vc ' vpk cos (Tt % * % 2B/3). (3.43) vd ' vpk cos (Tt ! 2 % *); vq ' vpk sin (Tt ! 2 % *). (3.44) vd ' vpkcos *; vq ' vpksin *. (3.45)
where p2s
is the slip velocity in elec rad/s, i.e. the angular velocity of the d,q axes relative to a point
fixed on the rotor. Thus in synchronously rotating axes p2s= T!T
r, with T = 2B f, but in d,q axes fixed
to the rotor, p2s
= 0. Note that R
D= R
Q= R
A.The electromagnetic torque is determined from the power associated with the stator current and the speed voltages, divided by the angular velocity in mechanical rad/s. If P is the number of poles,
The dynamical equation of motion is
where Tr
is in elec rad/s and T
m = Tr × 2/P is the rotor speed in rad/s. D is a viscous damping term.Transient simulation
Transients are calculated by integrating the differential eqns. (3.38),(3.39) and (3.40) in the form.
where T is the velocity of the rotating reference frame in elec rad/s. At the end of each timestep, new values of Rd
, R
q, R
D, R
Q, and T
rbecome available and the currents must be updated by solving eqns.
(3.34!3.36). Likewise the torque must be updated by means of eqn. (3.40). A typical timestep would be 0.002 s, but it depends on the time-constants of the particular motor. The system of equations is non- linear because of products like Rdiq
in the torque equation and T R
din the voltage equation.
If the d,q axes are fixed to the rotor, the speed voltages in eqns. (3.42) vanish, since T = Tr
. The
integration is "driven" by the applied voltages vd
and v
q, which are defined in terms of v
a,
vb, and v
c.
while the flux-linkages and currents are "outputs". For example, suppose the supply is defined by
where T = 2B f and * is an arbitrary phase angle. According to the transformation equations
If 2 = 2r the d,q axes are fixed to the rotor. If 2 = Tt they rotate at synchronous speed, with T = 2B f.
vab ' vpk 3 cos (Tt % B/6); vbc ' vpk 3 cos (Tt ! B/2); vca ' vpk 3 cos (Tt % 5B/6). (3.46) LLL ' Laa % Lbb ! 2Lab ' 2L11. (3.47) id ' 2
3ia [cos 2 ! cos (2 ! 2B/3)] ' 2ia/ 3 cos (2 % B/6);
iq ' !2
3ia [sin 2 ! sin (2 ! 2B/3)] ' !2ia/ 3 sin (2 % B/6)
(3.48)
pRD ' !RDiD. (3.49)
Fig. 3.17 Non-simultaneous switching of the supply phases
Non-simultaneous switching of the phases
Fig. 3.17 shows the connection of the motor to the supply. The contacts in line a are assumed to be already closed, with no current flowing. Those in line b close at the "point on wave" Tt = 21, and line
c closes at the angle Tt = 21 + 22; that is, after a delay 22/T measured from the closure of line b.
With line-neutral voltages given by eqns. (3.43), then with * = 0 the line-line voltages are
During the interval 21 # Tt # 22 vc
is undefined, but i
c= 0 and i
a= !i
b. Current flows in the loop ab. No
torque is produced, and the rotor remains stationary. A particular case of interest is to close line b at a peak of vab, that is, when Tt = !B/6, and then close line c after a further delay of 90E (2
1 = 0; 22 = 90E).This strategy eliminates most of the oscillatory component in the transient torque, [Wood, 1965]. During the delay, stator current flows only in the loop through phases a and b. The line-line resistance is RLL
= 2R
aand the line-line inductance L
LLis
Obviously current is induced in the rotor. The rotor circuit can be reduced to a single circuit by fixing the d-axis to the rotor and aligning it with the axis of the effective stator coil ab that results from the loop current flowing in lines a and b. Then the whole circuit is reduced to that of a single-phase transformer. With ic
= 0 we have
and the required alignment is achieved by setting 2 = 2r = !B/6, with id
=2 i
a/%3 and iq= 0: that is, all
the stator current is in the d-axis and none in the q-axis.
On the rotor, the Q coil is orthogonal to the axis of the effective stator coil ab and so RQ
= i
Q= 0, while
eqn. (3.42) holds for RD
with T = T
r= 0:
The mutual inductance between the rotor D-coil and the stator d-coil is L12 = 3/2 × LaA
. Since the d- and
a- coils have the same number of turns, the mutual inductance between the D-coil and phase a is also
RLL ' LLLia % 3L12iD RD ' L12id % LDiD (3.50) Rd ' L11id % L12iD; RD ' L12id % LDiD. (3.52) pRLL ' vLL ! 3 2 RLLid. (3.53)
Fig. 3.18 Alignment of dq axes with the conducting loop in phases a and b.
RLL ' Rab ' Ra ! Rb ' 3 Rd (3.51)
L12 cos (!B/6) = %3/2 × LaA
. The axis of phase b is at an angle 5B/6 relative to the d-axis, so that with
current in the stator loop ab and ib = !ia phase b makes an equal contribution to Rd
. Therefore the total
mutual inductance MD-LL
is 2 × L
12× %3/2 = %3 L
12, and
where RLL is the flux-linkage of the series connection of phases ab. Using the inverse reference-frame transformation it can be shown that
and since ia = /3/2 id and LLL = 2 L11, we can write eqns. (3.50) as
This agrees with eqns. (3.34) and (3.36), since LD = L22 . The voltage equation for the stator loop is
The solution during the delay interval can therefore proceed by integrating eqns. (3.49) and (3.53), updating the currents at each timestep using eqns. (3.50); there is no torque and no rotation. After line
c closes, all three motor terminal voltages are known: eqn. (3.45) can be used for vd
and v
q, and the
solution can proceed in d,q axes as before, provided that the final currents at the end of the delay are transformed into initial values for id
and i
q.vd ' Rdid ! TRq ' Rdid ! T(L11iq % L12iQ);
vq ' Rqiq % TRd ' Rqiq % T(L11id % L12iD). (3.54)
ia ' [id cos 2 ! iq sin 2] ' id cos Tt % iq cos (Tt % B/2). (3.55)
Vd % j Vq ' [R1 % j TL11] (Id % j Iq) % j TL12(ID % j IQ) (3.56) 0 ' [R2 % j sTL22] (ID % j IQ) % j sTL12(Id % j Iq). (3.57) a51 ' 3 2 (P/2)2 J [ iq0 ! Rq0 LsN] (3.59) a52 ' ! 3 2 (P/2)2 J [ id0 ! Rd0 LsN] (3.60) Steady-state operation
The theory so far assumes that we know the inductances L11
,
L12etc., but we need to relate these to the
familiar "equivalent circuit" parameters of the induction motor: R1
,
R2,
X1,
X2, and X
m. In the steadystate the terms in eqns. (3.42) containing the p operator are zero so that
We can combine these equations by writing V = Vd
+ jV
qwhere V
d= v
d/%2 and Vq= v
q/%2 and similarlywith the currents. This can be seen as follows: if 2 = Tt then From eqn. (3.54) it follows that
where R1
= R
aand L
12= 3/2 × L
aA. We can write I
2= I
D+ j I
Qdirectly for the "rotor current", while
for the "stator current" I1= I
d + jIq. The construction of the equivalent circuit requires also the
equivalent of eqn. (3.56) for the rotor, i.e.
where R2
= R
A= R
D. Eqns. (3.56) and (3.57) represent the conventional equivalent circuit and so we
have done enough to show that Rd
= R
q= R
a;
L11= [X
1+ X
m]/T,
L22= [X
2+ X
m]/T, and L
12= 3/2 × L
aA= Xm
/T. Note the appearance of slip s in eqn. (3.57): by dividing this equation throughout by s, the
rotor circuit is "referred" to the stator with the same frequency T and the familiar R2/s appears as the
only manifestation of rotation.
Small-signal analysis
For this we assume small perturbations (denoted by )) about a steady operating point (denoted by subscript 0): vd
= v
d0+ )v
d; vq= v
q0+ )v
q; id= i
d0+ )i
d; iq= i
q0+ )i
q; iD= i
D0+ )i
D; iQ= i
Q0+ )i
Q; Rd =Rd0 + )Rd; Rq = Rq0 + )Rq; RD = RD0
+ )R
D; RQ= R
Q0 + )RQ; Tr= T
r0 + )Tr; and Te = Te0 + )Te. T is missingbecause it is considered fixed and equal to 2Bf, but Tr
can vary. If we substitute these into the voltage
eqn. (3.42) and the torque eqn. (3.40) and simplify by subtracting the "steady-state" components, while ignoring products of the form )x)y, then after some grinding we getp)Rd !1/JsN T kr/JsN 0 0 )Rd )vd p)Rq !Tr0 !1/JsN 0 kr/JsN 0 )Rq )vq p)RD = ks/JrN 0 !1/JrN T ! Tr0 !Rq0 )RD + 0 (73) p)RQ 0 ks/JrN 0 !1/JrN Rd0 )RQ 0 p)Tr a51 a52 a53 a54 !D/J )Tr 0 where
a53 ' 3 2 (P/2)2 J kr LsNRq0 (3.61) a54 ' ! 3 2 (P/2)2 J kr LsNRd0 (3.62) vd0 ' Rdid0 ! TRq0; Rq0 ' L11iq0 vq0 ' Rqiq0 % TRd0; Rd0 ' L11id0 (3.63) id0 ' TL11vq0 % Ravd0 Ra2 % T2L 112 ; iq0 ' ! TL11vd0 ! Ravq0 Ra2 % T2L 112 (3.64)
Fig. 3.19 Equivalent circuit of split-phase motor. During start-up, the cut-out switch switches from "start" to "run" at a certain percentage of the synchronous speed. The OC (open-circuit) position of the cut-out switch simply represents here the electrical connections of a single-phase motor.
and JsN = LsN/R1, Jr
N = L
rN/R2, LsN = (LsLr ! M2)/Lr,
LrN = (LsLr ! M2)/Ls, ks = M/Ls, kr = M/Lr, and P is
the number of poles. The notation has been changed using Ls = L11; Lr = L22, and M = L12 so that the matrix eqn. (3.58) can be more easily compared with Vas' eqn. 2.10-36 [9] which is derived using space- vector theory.
In the calculation of the coefficients a51..a54, the determination of Rd0 and Rq0 generally requires the
inversion of a 4 × 4 matrix but this can be avoided if the eigenvalues are calculated for the zero-slip condition. This makes iD0 = iQ0
= 0 and T!T
r0 = 0. From eqn. (3.57) the steady-state conditions arefor which the solution is