You are on page 1of 13

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON ENERGY CONVERSION 1

Constant-Parameter Voltage-Behind-Reactance
Induction Machine Model Including
Main Flux Saturation
Francis Therrien, Student Member, IEEE, Mehrdad Chapariha, Student Member, IEEE,
and Juri Jatskevich, Senior Member, IEEE

Abstract—Interfacing of electrical machine models with ac saturation characteristics [12], [14]. Consequently, many simu-
power networks has a significant impact on numerical accuracy lation programs assume constant leakage inductances (e.g., see
and efficiency in state-variable-based transient simulation pro- [1] and [2]). To develop a straightforward-to-use model, this
grams. This paper continues the recent work in this area by propos-
ing a new explicit constant-parameter voltage-behind-reactance paper considers only saturation of the main flux, which offers
(VBR) induction machine model that includes main flux satura- sufficient accuracy for many practical cases and system-level
tion and allows a direct interface to any external network. The transient studies [5].
proposed model uses numerical approximations to achieve a de- In transient simulation programs [1]–[4], induction machines
coupled interfacing circuit with constant RL branches that is easy are typically modeled in a rotating qd-reference frame using
to use in many simulation programs. Computer studies demon-
strate that the proposed model provides high numerical accuracy transformed variables [5], [15], [16]. The classical qd models
for machines with a diverse range of parameters, even at fairly offer many numerical advantages including the absence of rotor-
large integration step sizes. position-dependent inductances, magnetically decoupled q- and
Index Terms—Induction machine, interfacing circuit, magnetic
d-axis circuits, and constant qd-variables in steady state [5].
saturation, state-variable-based program, transient simulation, This paper considers state-variable-based transient simulation
voltage-behind-reactance (VBR) model. programs (e.g., see [1] and [2]), wherein the built-in qd models
are typically interfaced with external circuits using controlled
current sources [16]. Therefore, when a machine is in series
I. INTRODUCTION
with an inductive or ideal switching component (which is very
NDUCTION machines normally operate under lightly-to- common in practical systems), a proper state model cannot be
I moderately saturated conditions. As machine saturation is a
complex phenomenon, refined analysis techniques such as finite
generated due to incompatible input–output interfacing [16].
To achieve a compatible interface, fictitious snubber circuits
element and/or coupled magnetic circuit methods may be used are typically used, which introduce additional error and lead to
during the design stage when very precise results are required increased numerical stiffness and longer CPU times [16].
and the computational cost and time are of a lesser priority. For The so-called coupled-circuit phase-domain (CC-PD) models
power system transient studies, lower order general-purpose [16], [17] are represented in physical abc-variables (phase coor-
lumped-parameter models are typically preferred due to their dinates), therefore allowing direct interfacing with any external
relative simplicity and high numerical efficiency. In such mod- circuit. However, the CC-PD models remain computationally
els, magnetic nonlinearities are often represented by saturating expensive due to their fully coupled rotor-position-dependent
the fundamental component of the main magnetizing flux [1]– inductance matrices [18], [19]. From a numerical perspective,
[10]. In addition to increasing the overall modeling accuracy, the voltage-behind-reactance (VBR) models [18]–[25] appear
adequate representation of main flux saturation is crucial for to be a better alternative. In these models, the stator interfacing
many indirect rotor field-oriented control schemes [5], [11]. circuit is in abc-phase coordinates, allowing a direct interface
Saturation of leakage inductances can also be used to further in- with external networks, while the rotor subsystem is represented
crease model fidelity [7], [9], [12], [13], particularly for uncon- in qd-coordinates, yielding better scaled eigenvalues and higher
trolled motor start-ups. However, unlike the main flux saturation numerical efficiency [18].
characteristic which can be easily obtained from a no-load test, It is emphasized that under the same set of assumptions (e.g.,
complicated procedures are required to determine the leakage representation of saturation or not), the qd, CC-PD, and VBR
models are algebraically equivalent. In other words, these mod-
Manuscript received October 8, 2013; revised April 11, 2014; accepted els yield identical analytical solutions (when converted to a com-
July 3, 2014. This work was supported by the Natural Science and Engineer- mon reference frame). However, they have different numerical
ing Research Council (NSERC) of Canada under the Discovery Grant. Paper properties (e.g., interfacing circuits and eigenstructures), which
no. TEC-00596-2013.
The authors are with the Department of Electrical and Computer Engineering, can have a considerable impact on numerical efficiency and ac-
University of British Columbia, Vancouver, BC V6T 1Z4, Canada (e-mail: curacy when these models are interfaced with power networks
francist@ece.ubc.ca; mehrdadc@ece.ubc.ca; jurij@ece.ubc.ca). inside simulation programs.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. Due to dynamic saliency, the interfacing circuit of the original
Digital Object Identifier 10.1109/TEC.2014.2342258 magnetically linear VBR synchronous machine model contains

0885-8969 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON ENERGY CONVERSION

variable inductances [18]. While being more efficient than CC- II. QD MODEL
PD models [18], it is nonetheless computationally expensive
To set the stage for the derivation of the new model, it is
due to the need for refactorization (or updating) of the state useful to recall the equations of the classical general-purpose
matrices (equations) [20]. Several techniques have since been
lumped-parameter symmetrical induction machine model in qd-
proposed to accurately neglect dynamic saliency and achieve a
coordinates [5]. Herein, the motor sign convention and the arbi-
constant-parameter VBR synchronous machine model: adding trary reference frame are used. All parameters are referred to the
a fictitious high-frequency damper winding [20], applying sin-
stator side. To incorporate the double-cage or deep-bar effects
gular perturbation theory [21], and transferring the time-varying (if desired) [26], the induction machine is assumed to have N
inductances of the interfacing circuit into its subtransient voltage equivalent rotor windings with no end-ring common resistance.
source [22]. In particular, the numerical efficiency of the latter
All windings on the same magnetic axis are assumed to have
approach can be significantly improved by using continuous- identical coupling. If necessary, standard double-cage models
or discrete-time filters to form a proper and explicit state-space with Llr 1 = 0 and nonidentical coupling [10] can be converted
model [22]. It is stressed that all of the above techniques/models
to equivalent models with identical coupling by following the
[20]–[22] are based on the assumption of magnetic linearity. In- procedure set forth in [27]. The corresponding voltage equations
corporation of main flux saturation into these models would are [5]
cause their interfacing circuit inductances to become time-
dependent. vq s = rs iq s + ωλds + pλq s (1)
Unlike the synchronous machine model [18], the magnet-
ically linear VBR induction machine model has a constant- vds = rs ids − ωλq s + pλds (2)
parameter interfacing circuit since its rotor is isotropic [19]. v0s = rs i0s + pλ0s (3)
However, the saturable VBR induction machine models [1],
[23] (which are based on the synchronous machine model pre- vq r j = rr j iq r j + (ω − ωr )λdr j + pλq r j , j = 1, . . . , N (4)
sented in [24]) have time-varying interfacing circuits. Con- vdr j = rr j idr j − (ω − ωr )λq r j + pλdr j , j = 1, . . . , N (5)
sequently, these models also suffer from poor numerical
efficiency. v0r j = rr j i0r j + pλ0r j , j = 1, . . . , N . (6)
The main objective of this paper is to propose a numeri-
cally accurate and efficient constant-parameter VBR induction The flux linkage equations are
machine model including main flux saturation. This paper is a
continuation of the prior work in this area [18]–[25]. The con- λq s = Lls iq s + λm q (7)
tributions of this paper and the properties of the new model are λds = Lls ids + λm d (8)
summarized as follows.
1) A constant-parameter VBR induction machine model in- λ0s = Lls i0s (9)
cluding main flux saturation is derived. To the best of λq r j = Llr j iq r j + λm q , j = 1, . . . , N (10)
our knowledge, it is the first time that a saturable ac ma-
chine model with a constant-parameter interfacing circuit λdr j = Llr j idr j + λm d , j = 1, . . . , N (11)
is proposed. λ0r j = Llr j i0r j , j = 1, . . . , N (12)
2) The new model uses the unified decoupled RL branch
stator interfacing circuit presented in [25]. where the magnetizing fluxes are given by
3) To obtain an efficient and explicit formulation, the pro-
⎛ ⎞
posed model expands the numerical approximation tech- 
N
niques originally proposed in [22] to take into account sat- λm q = Lm (λm )im q = Lm (λm ) ⎝iq s + iq r j ⎠ (13)
uration in both magnetic axes (including cross-saturation). j =1
4) The numerical error introduced by the proposed approx- ⎛ ⎞
imation is analyzed and related to the integration step 
N

size, the machine parameters, and the saturation level. λm d = Lm (λm )im d = Lm (λm ) ⎝ids + idr j ⎠ . (14)
Several machines with diverse range of parameters are j =1

considered to investigate the applicability of the proposed


approach. Here, Lm (λm ) denotes the saturation-dependent steady-state
5) The presented computer studies demonstrate that the pro- magnetizing inductance, which can be found from the machine
posed model possesses high numerical efficiency and ac- no-load saturation characteristic [6]
curacy for machines with diverse parameters even at fairly
large integration step sizes. λm = F −1 (im ). (15)
6) It is envisioned that the proposed model will find its ap-
plication in many commercial-grade transient simulation Since the machine is round and symmetrical, the main mag-
programs and packages, wherein such models can be used netizing flux λm and current im are defined as
as built-in components readily available to thousands of  
engineers and users. λm = λ2m q + λ2m d and im = i2m q + i2m d . (16)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 3

Finally, the developed electromagnetic torque and mechanical Substituting (21) and (22) into (19) consecutively and solving
equation of single rigid-body motion are given by for pλm q d yields the following state equations:
⎛ ⎞
3P

 N

Te = (λm d iq s − λm q ids ) (17) λm q i p λ
4 p = Lm i (λm q d ) ⎝p q s + qrj ⎠
 λm d ids Llr j λdr j j =1
2
Te − T m = J pωr (18) (23)
P
where
where P is the number of poles and J denotes the combined   ⎛ ⎞−1
Lm q q Lm q d  N
1
rotor-load inertia. Lm i (λm q d )≡ =⎝Γm i (λm q d )+ I2⎠
Lm q d Lm dd j =1
Llr j
III. VARIABLE-PARAMETER VBR MODEL (24)
The derivation of the saturable variable-parameter VBR in- and In is the n-by-n identity matrix.
duction machine model (VP-VBR) has been presented in [23]. To formulate a standard state-space model, (23) must also
The main steps of this derivation are briefly explained here. be rewritten as a function of state and input variables. First,
For magnetically linear VBR models, the state variables of the inserting (7), (8), and (23) into (1) and (2) and solving for piq ds
equivalent rotor subsystem are typically chosen to be the rotor gives


−1
flux linkages [18], [19]. When modeling main flux saturation, iq s Lls + Lm q q Lm q d
it is convenient to also consider the magnetizing fluxes λm q p
ids
=
Lm q d Lls + Lm dd
and λm d as state variables [28]. Consequently, one pair of q- 



and d-axis rotor flux linkages is removed from the set of state vq s rs ωLls iq s
× −
variables. Without loss of generality, in this paper, λq r 1 and λdr 1 vds −ωLls rs ids
are excluded from the state vector and instead are considered


λm d e
auxiliary algebraic variables. −ω − q (25)
−λm q ed
Defining Γm (λm ) ≡ 1/Lm (λm ), differentiating (13) and
(14) with respect to time, and using the chain and quotient where
⎡ ⎤ ⎡ ⎤⎛ ⎞
rules yield eq Lm q q Lm q d  N



⎣ ⎦=⎣ ⎦⎝ p λq r j ⎠
λm q . (26)
im q
Lm q d Lm dd Llr j λdr j
p = Γm i (λm q d )p (19) ed j =1
im d λm d
Next, solving (4) and (5) for pλq dr j yields
where



⎡ λ vq r j i
dim λ2m q p qrj = − rr j q r j
− Γm (λm ) + Γm (λm ) λdr j vdr j idr j
⎢ λ2m
⎢ dλm

Γm i (λm q d ) = ⎢
⎢ −(ω − ωr )
λdr j
⎣ dim −λq r j
, j = 1, . . . , N (27)
λm q λm d
− Γm (λm )
dλm λ2m where iq dr j is defined using (22). Due to the choice of state
 ⎤ variables, (27) is only integrated for j = 2, . . . , N . Finally, in
dim λm q λm d
− Γm (λm ) ⎥ order to obtain an explicit model, λq dr 1 must be written in terms
dλm λ2m ⎥
⎥ (20) of state and input variables. Substituting (22) into (13) and (14)
 ⎥. and solving for λq dr 1 yields
dim λ2m d ⎦
− Γm (λm ) + Γm (λm ) ⎛⎛ ⎞
dλm λ2m
N

λq r 1 1 ⎠ λm q
= Llr 1 ⎝⎝Γm (λm ) +
Based on (13) and (14), the left-hand side of (19) is explicitly λdr 1 Llr j λm d
j =2
defined as ⎞






 N
iq s
N
1 λq r j ⎠ λ
i i i − − + m q . (28)
p mq = p qs + p qrj . (21) ids L λdr j λm d
im d ids idr j j =2 lr j
j =1
The next step is to define the equation for the stator interfacing
The rotor currents in (21) must be written as a function of circuit. Inserting (7), (8), and (23) into (1) and (2) yields
state and input variables. As a first step, the rotor currents can

 
be expressed in terms of rotor flux linkages and magnetizing vq s iq s −L m q d L l s + L m q q
= rs +ω
fluxes by algebraically manipulating (10) and (11), giving vd s id s −L l s − L m d d L m q d

 
  



iq s L l s + L m q q L m q d

iq s eq
iq r j 1 λq r j λ × + p +
= − mq , j = 1, . . . , N. id s L m q d L l s + L m d d id s e
idr j Llr j λdr j λm d d

(22) (29)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON ENERGY CONVERSION

where IV. CONSTANT-PARAMETER VBR MODEL


 


eq Lm q d −Lm q q To obtain the desired constant-parameter (CP-VBR) model,
iq s
= ω it is necessary to satisfy conditions (35)–(37) at all times. This
ed Lm dd −Lm q d ids
is achieved if Lm i (λm q d ) becomes a constant scalar matrix, i.e.,


a diagonal matrix with identical nonzeros.
λm d e
+ω + q . (30)
−λm q ed
A. Derivation of Constant-Parameter Interfacing Circuit
Adding zero-sequence [(3) and (9)] to (29) and applying the The first step consists of substituting (21) into (19) and sepa-
inverse Park’s transformation [5] yield the machine–network rating Γm i (λm q d ) into constant and time-dependent terms
interfacing equation

 N


iq s iq r j λm q
vabcs = rs iabcs + (Lls I3 + Ls (θr ))piabcs + eabcs (31) p + p = Γm u p
ids idr j λm d
j =1
where fabcs = [ fas fbs fcs ]T ( f may represent v or i), eabcs =

  λ
K−1 T
s [ eq ed 0 ] , and Ks is Park’s transformation matrix [5]. + ΔΓm is (λm q d )p m q (38)
λm d
The subtransient inductance matrix Ls (θr ) is defined as
⎡ ⎤ where Γm u ≡ 1/Lm u is the unsaturated magnetizing inverse
1 −0.5 −0.5 inductance, and
Ls (θr ) = La ⎣ −0.5 1 −0.5 ⎦
−0.5 −0.5 1 ΔΓm is (λm q d ) =
⎡ ⎤ ⎡ 2
dim λm q
L(θr ) L(θr + 120◦ ) L(θr − 120◦ ) − Γ (λ ) + ΔΓm (λm )
⎢ ⎥ ⎢ dλm m m
λ2m
+ ⎣ L(θr + 120◦ ) L(θr − 120◦ ) L(θr ) ⎦ (32) ⎢

◦ ◦ ⎢
L(θr − 120 ) L(θr ) L(θr + 120 ) ⎣ dim λm q λm d
− Γm (λm )
where the function L(ϕ) is dλm λ2m
 ⎤
dim λm q λm d
L(ϕ) = Lb cos(2ϕ) + Lc sin(2ϕ) (33) − Γm (λm ) ⎥
dλm λ2m ⎥
⎥ (39)
and ϕ = {θr , θr − 120◦ , θr + 120◦ }. The equivalent induc-  2 ⎥
dim λm d ⎦
tances La , Lb , and Lc are defined as − Γm (λm ) + ΔΓ m (λ m )
dλm λ2m
Lm q q + Lm dd Lm q q − Lm dd 2
La = , Lb = , Lc = Lm q d . where ΔΓm (λm ) = Γm (λm ) − Γm u . Substituting (22) into
3 3 3 (38) and solving for the pλm q d vectors multiplied by time-
(34)
The resulting interfacing circuit (31) has a constant diagonal independent elements yields
resistance matrix; however, its inductance matrix is full and



λm q iq s  N
p λq r j

dependent on the rotor angle θr . From (32)–(34), it can be p = Lm u p +
λm d ids L λdr j
observed that Ls (θr ) will become constant if the following j =1 lr j
three conditions are satisfied:

λm q
− ΔΓm is (λm q d )p (40)
Lm q q + Lm dd = κ = const. ∀t (35) λm d
Lm q q − Lm dd = 0 ∀t (36) where
Lm q d = 0 ∀t (37)  −1  N
1 1 1
Lm u = Γm u + and = . (41)
where κ is an arbitrary time-independent scalar. Conditions LΣlr LΣlr L
j =1 lr j
(35)–(37) are observed in the magnetically linear symmetri-
Substituting (23) into the right-hand side of (40) yields
cal induction machine model [19]. However, when saturation 

is considered, it can be seen from (20) and (24) that the three
 N

λm q  iq s p λq r j
conditions are not always met. p = Lm u p +
λm d ids L λdr j
It is important to keep in mind that the saturable VP-VBR j =1 lr j


model derived in this section is algebraically equivalent to
i
the general-purpose lumped-parameter qd model defined in −ΔΓm is (λm q d )Lm i (λm q d ) p q s
ids
Section II, i.e., it is possible to convert one model into the other
using algebraic manipulations. However, unlike the qd model,  N

the VP-VBR model can be interfaced directly to arbitrary exter- p λq r j
+ . (42)
nal networks. L λdr j
j =1 lr j
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 5

Inserting (7), (8), and (42) into (1) and (2) then gives voltage source eabcs . As a consequence, the term eq d is a func-



tion of the current derivative piq ds .
vq s i ids
= rs q s + ω(Lls + Lm u ) One method to calculate piq ds in (44) is to rewrite it in terms of
vds ids −iq s
state and input variables using (25), as is done in (23). However,

   since eq d is an output of the equivalent rotor subsystem, this
i eq
+(Lls + Lm u )p q s +  (43) approach results in an algebraic feedthrough between eq d and
ids ed
vq ds . In the case where vq ds is an output of the external circuit
where subsystem (e.g., if the machine is connected to an inductive
 


branch), vq ds will be algebraically related to eq d . Consequently,
eq λm d −ids
N
p λq r j
=ω 
+ Lm u ω + the state model will have an algebraic loop containing eq d and
ed −λm q iq s Llr j λdr j vq ds [22]. Models with algebraic loops generally require special
j =1

iterative solvers which are not available in some simulation
 i programs (e.g., see [1]), reduce simulation robustness [29], and
−ΔΓm is (λm q d )Lm i (λm q d ) p q s
ids increase the overall computational time [6], [30].


N
 An alternative approach is to approximate piq ds in (44) using
p λq r j numerical techniques. Specifically, piq ds is herein approximated
+ . (44)
Llr j λdr j
j =1
using backward differentiation formulas (BDFs) [22], [31]. A
similar approach was originally presented in [22] to achieve
Combining (43) with (3) and (9) and applying the in- an explicit constant-parameter magnetically linear VBR syn-
verse Park’s transformation results in the following constant- chronous machine model. Therein, due to dynamic saliency,
parameter machine–network interfacing equation: only piq s needed to be approximated; in addition, piq s was mul-
vabcs = rs iabcs + (Lls I3 + Lcs )piabcs + eabcs (45) tiplied by a constant value [22]. This is in contrast to the pro-
posed model, wherein due to saturation (and cross-saturation),
where the new constant subtransient inductance matrix is piq ds is multiplied by a time-dependent coupled matrix [see
⎡ ⎤
1 −0.5 −0.5 (44)].
2
Lcs = Lm u ⎣ −0.5 1 −0.5 ⎦ . (46) Assuming a variable-step solver for generality, the first-order
3
−0.5 −0.5 1 backward differentiation formula (BDF1) is

The constant-parameter VBR model (45) has been achieved (k ) (k −1)


(k ) iq ds − iq ds
without modifying the rotor subsystem state equations (23) and pĩq ds = (51)
Δt(k )
(27).
The final step consists of modifying (45) to achieve an inter- where the superscript (k) indicates the value at the kth time step,
facing circuit comprised only of decoupled RL branch elements and Δt(k ) = t(k ) − t(k −1) . Here, the tilde symbol () denotes
[25]. Assuming a wye-connected (grounded or floating neutral) an approximation as opposed to the exact value. If a more ac-
machine, the terminal stator voltage vabcs is decomposed into curate approximation is desired, the second-order BDF (BDF2)
two vectors by following the procedure set forth in [25] can also be used
vabcs = vabcn + vn g (47)   1 1

(k ) (k ) (k −1)
pĩq ds = iq ds − iq ds +
where Δt(k ) Δt(k ) + Δt(k −1)
 (k −1) (k −2)

vabcn = rs iabcs + LD piabcs + eabcs (48) Δt(k ) iq ds − iq ds
− (k −1) . (52)
T Δt Δt(k ) + Δt(k −1)
and vn g = [ vn g vn g vn g ] . The voltage vn g is defined as
vn g = L0 pin g (49) Replacing piq ds in (44) by (51) or (52), an explicit constant-
parameter VBR induction machine model with main flux satu-
where in g = ias + ibs + ics . Finally, the newly introduced in-
ration is obtained. A diagram depicting the implementation of
ductances LD and L0 are related to Lls and Lm u by
this new model is shown in Fig. 1, wherein the interfacing cir-
Lm u cuit defined by (47)–(50) is contained within the shaded area.
LD = Lls + Lm u and L0 = −. (50)
3 In general, other approximations (e.g., see [22]) could also be
Equations (47)–(50) define the decoupled RL branch inter- used instead of (51) or (52).
facing circuit [25, Fig. 1]. Up to this point, the CP-VBR model
is also algebraically equivalent to the qd and VP-VBR models. V. ERROR ANALYSIS
If the solver uses a high-order integration rule and the integra-
B. Numerical Approximation tion step size is sufficiently small, it can be assumed that the nu-
The procedure presented in Section IV-A is analogous to merical error in the proposed model will mostly come from the
moving the time-varying elements of the algebraically exact approximation of Lm u ΔΓm is (λm q d )Lm i (λm q d )piq ds in (44)
inductance matrix Ls (θr ) defined in (32) into the subtransient using BDFs [(51) or (52)]. This overall error is thus determined
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON ENERGY CONVERSION

TABLE I
COEFFICIENTS L m u /L Σ l r AND χ −1 FOR VARIOUS INDUCTION MACHINES

Ref. [5] [9] [5] [5] [5] [10] [27] [27] [27]

Hp 3 5 50 500 2250 10 14.7 121 845


No. of rotor windings 1 1 1 1 1 2 2 2 2
L m u /L Σ l r 34.7 14.3 43.3 44.8 57.7 214 52.5 72.4 94.5
χ −1 72.3 31.6 89.6 92.6 118 999 135 200 285

ac network by the program solver. However, the equation re-


lated to Lm u ΔΓm is (λm q d )Lm i (λm q d ) is solved inside the ro-
tor subsystem using BDFs, which introduce additional numer-
ical error [see (55) and (56)]. Therefore, smaller values of
Fig. 1. Constant-parameter VBR (CP-VBR) formulation of the saturable in-
Lm u ΔΓm is (λm q d )Lm i (λm q d ) should result in smaller over-
duction machine model. all error, as the effect of the approximated saturation-dependent
term will be minimized with respect to the remaining constant
inductance Lls + Lm u . The ratio of these two inductances is
by two major factors: 1) the BDFs, and 2) the magnitude of the therefore considered as an indication of how accurate the pro-
saturation-dependent matrix Lm u ΔΓm is (λm q d )Lm i (λm q d ). posed approximation is expected to be
Lm u ΔΓm is (λm q d ) L m i (λm q d )2
A. Numerical Differentiation Error ζ(λm q d ) = . (57)
Lls + Lm u
The error associated with the BDFs can be related to the To get more insight into the expected model accuracy, it is
integration step size. Assuming a constant step size Δt, the particularly useful to understand how ζ(λm q d ) depends on ma-
(k )
local truncation error (LTE) dBDF1 [31] of BDF1 is defined as chine parameters. This analysis can be simplified by neglecting
(k ) (k −1) cross-saturation in (57). The reasoning behind this assumption
(k ) iq ds − iq ds (k )
dBDF1 = − piq ds . (53) is presented in [32], wherein it is shown that cross-saturation has
Δt a marginal effect on the accuracy of induction machine models
Similarly, the LTE of BDF2 is given by with stator currents and rotor and/or magnetizing fluxes as state
2  (k )  1  (k )  variables. Therefore, equating dim /λm to Γm (λm ) in (57) [32],
(k ) (k −1) (k −2) (k )
dBDF2 = iq ds − iq ds − iq ds − iq ds − piq ds . after some algebraic manipulations, ζ(λm q d ) can be rewritten
Δt 2Δt
(54) as
Inserting Taylor series in (53) and (54), after some algebraic ζ(λm ) = χ · Ksat (λm ) (58)
manipulations, we obtain
where the saturation-independent scalar χ and the saturation-
(k ) Δt 2 (k )
dBDF1 = − p iq ds + O(Δt2 ) (55) dependent function Ksat (λm ) are
2
Δt2 3 (k ) Lm u LΣlr
(k )
dBDF2 = − p iq ds + O(Δt3 ). (56) χ= · (59)
3 Lls + Lm u LΣlr + Lm u

The error introduced by the BDF approximations is thus pro- Lm (λm ) LΣlr + Lm u
Ksat (λm ) = 1 − . (60)
portional to the step size Δt for BDF1 and Δt2 for BDF2, Lm u LΣlr + Lm (λm )
which also proves the consistency [31] of these approaches.
Based on (59) and (60), it becomes apparent that ζ(λm ) de-
Moreover, assuming a purely sinusoidal signal and using the
pends on both the machine parameters and the saturation level.
synchronous reference frame, it can be seen that (51) and (52)
In particular, since for induction machines Lm u  LΣlr , we can
will not introduce any error in steady state. Finally, the error
assume that Ksat (λm ) mostly depends on the saturation level,
will be maximized in situations where the second and/or third
and thus that the effect of machine parameters is essentially
derivatives of iq ds are large, for example, during fast and/or
captured by χ. It can also be seen from (60) that for unsaturated
large disturbances.
conditions Ksat (λm ) = 0 and that Ksat (λm ) monotonically in-
creases as a function of the level of saturation. Finally, since Lls
B. Approximation Error and Machine Parameters
and Lm u typically have the same order of magnitude, the value
Substituting (44) into (43), it is observed that the term of χ is predominantly defined by the simple ratio Lm u /LΣlr .
Lm u ΔΓm is (λm q d )Lm i (λm q d ) is a saturation-dependent induc- The values of Lm u /LΣlr and χ−1 for several machines with
tance matrix in series with the constant inductance Lls + Lm u . a diverse range of parameters and power ratings are summa-
The differential equation related to Lls + Lm u is converted rized in Table I. As this table shows, Lm u /LΣlr and χ−1 are
to abc-coordinates and is solved together with the rest of the fairly correlated and generally increase with machine size. These
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 7

The network of Fig. 2 has been implemented using PLECS’


built-in circuit components. Since the motor is in series with an
inductive cable, only the VBR models can be interfaced directly
to the external network among the test case models [16]. The
qd model is represented using controlled current sources and
is interfaced using fictitious resistive snubbers Rsn [16]. To
Fig. 2. Example network used for the voltage sag transient study.
demonstrate the effect of snubbers, two values (large and small)
of Rsn have been considered. The corresponding models are
referred to as qd1 (which is numerically stiff) and qd2 (which
coefficients are also typically larger for machines modeled with
is less stiff), respectively.
two rotor windings instead of one. Finally, based on (58)–(60)
A reference solution is obtained by solving the saturable
and the magnitude of these coefficients in Table I, it can be
VP-VBR model using Simulink’s general-purpose solver ode45
concluded that ζ(λm ) should remain relatively small even for
with the following stringent settings: maximum and minimum
machines with less favorable parameters under heavy saturation.
step sizes of 10−5 and 10−8 s, respectively, and absolute and
This is a direct result and benefit of the VBR formulation which
relative error tolerances of 10−5 . Under these conditions, the
uses the subtransient inductance, wherein the magnetizing in-
numerical error is very small and the simulation results are
ductance Lm (λm ) appears in parallel with the typically much
assumed to be the reference solution. To achieve a fair compar-
smaller equivalent rotor leakage inductance LΣlr .
ison, due to the numerical stiffness of the qd1 model, all subject
models are simulated using the stiffly stable solver ode15s [29].
VI. COMPUTER STUDIES Typical settings are used: maximum and minimum step sizes of
The proposed CP-VBR model (see Section IV) and the al- 10−3 and 10−7 s, respectively, and absolute and relative error
gebraically exact VP-VBR model (see Section III) have been tolerances of 10−4 .
implemented in MATLAB/Simulink [29] using the PLECS tool- In this paper, the 2-norm (cumulative) relative error [34] of
box [1]. Two versions of the proposed model are considered. In the solution trajectories is adopted to quantify the models’ nu-
the first one, the current derivative piq ds in (44) is computed merical accuracy. For example, the 2-norm relative error of ias
using (51) (i.e., BDF1); in the second one, piq ds is computed is given by
using (52) (i.e., BDF2). The resulting formulations are referred  ref 
to as CP-VBR-BDF1 and CP-VBR-BDF2, respectively. To pro- ias − ias 
ε(ias ) = 2
× 100% (61)
vide a meaningful comparison, the classical qd model (with iref
as 2
winding flux linkages as the state variables) [5] is also imple-
mented, wherein saturation is represented using the explicit flux where iref
as and ias are the reference and calculated solution
correction method [6]. All machine models are simulated in trajectories of ias , respectively, for the duration of the considered
the synchronous reference frame. For consistency, all simula- transient study.
tions are executed on a PC with a 2.83-GHz Intel CPU running To verify the proposed approach, a few motors with different
Windows XP. parameters are considered.
In the considered case study, a low voltage (LV) wye-
connected induction motor is connected to a medium volt-
A. Induction Machine #1 (IM1)
age distribution (MV) network (represented by a three-phase
Thévenin equivalent) through a cable and a transformer. The The subject transient study is first executed using a 10-hp
single-line diagram of the network is shown in Fig. 2. The machine (IM1) with two rotor windings [10]. The machine pa-
motor is initially assumed to operate in steady state with ap- rameters and saturation curve are summarized in Appendix A.
plied base mechanical torque Tb . The source voltage Vs is set For this machine, the ratio of unsaturated magnetizing induc-
to 1.1 p.u. as to emulate a strongly saturated operation and tance over equivalent rotor leakage inductance Lm u /LΣlr (214)
thus increase the numerical error as explained in Section V-B. and the coefficient χ−1 (999) are very large (see Table I). There-
While this voltage is high for grid-connected induction motors, fore, as shown in Section V-B, the proposed CP-VBR models
it is not uncommon for machine-converter systems to operate at should yield excellent numerical accuracy. Snubber resistances
even higher voltages [33]. At t = 0.04 s, a single-line-to-ground Rsn of 250 Ω and 5 Ω are used in the qd1 and qd2 models,
fault is assumed to occur upstream in the network, resulting in respectively.
an unbalanced voltage sag in the test system. The voltage sag is The simulated stator current ias , rotor electrical speed ωr ,
modeled by decreasing the voltage on phase a of the equivalent electromagnetic torque Te , and main flux λm are shown in Fig. 3,
Thévenin source to 0.5 p.u. The fault is cleared 100 ms later, wherein the unbalanced conditions created by the single-phase
and Vs is restored back to 1.1 p.u. This study has been chosen voltage sag are apparent. Only phase a of the stator current is
because of the highly unbalanced currents and severe transients presented since it has the largest peaks during the disturbance.
occurring as a result of the voltage sag. Consequently, accord- To better visualize the performance of the models, a magnified
ing to Section V-A, the approximation error will be maximized view of the first peak of ias is also reproduced in Fig. 4. As
under such conditions, thus validating the model in its less ac- observed in Figs. 3 and 4, all models except qd2 accurately
curate operating region. predict the machine behavior in both steady-state and transient
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON ENERGY CONVERSION

Fig. 5. Saturation curve of IM1 and its operating range (denoted by the thick
line) during a single-phase voltage sag.

TABLE II
CALCULATED 2-NORM RELATIVE ERRORS OF ia s , ω r , T e , AND λm
FOR THE SINGLE-PHASE VOLTAGE SAG TRANSIENT (IM1)

Model VP-VBR CP-VBR- CP-VBR- qd1 (R s n = qd2


BDF1 BDF2 250Ω) (R s n = 5Ω)
Var.

i a s (%) 0.035 0.056 0.052 0.153 5.578


ω r (%) 0.001 0.002 0.002 0.004 0.208
T e (%) 0.033 0.060 0.055 0.205 7.895
λm (%) 0.004 0.010 0.009 0.090 4.445
Fig. 3. Transients in stator current ia s , rotor electrical speed ω r , electromag-
netic torque T e , and main flux λm during a single-phase voltage sag (IM1).

To quantify the accuracy achieved by the various models, the


2-norm (cumulative) relative errors of ias , ωr , Te , and λm with
respect to their reference trajectories are summarized in Table II.
As this table shows, all three VBR models yield errors below
0.1% for all variables, and no significant difference is observed
between CP-VBR-BDF1 and CP-VBR-BDF2. The qd1 model
is also highly accurate, as its 2-norm errors are all smaller than
0.21%. However, as observed in Figs. 3 and 4 and summarized
in Table II, the qd2 model is much less accurate with errors of
5.578% for ias and 7.895% for Te .

B. Induction Machine #2 (IM2)


Fig. 4. Magnified view of the first peak of stator current ia s during a single- The same voltage sag transient study is repeated here using a
phase voltage sag (IM1). 5-hp machine (IM2) [9]. The machine parameters are summa-
rized in Appendix B. For this machine, the very low values of
Lm u /LΣlr (14.3) and χ−1 (31.6) indicate that the proposed ap-
conditions. The inaccuracy of qd2 is caused by its fairly small proximation method should be less accurate. In this study, Rsn
snubber. is set to 300 Ω and 3 Ω for the qd1 and qd2 models, respectively.
In Section V-B, the level of saturation was found to affect the The predicted stator current ias , rotor electrical speed ωr ,
solution accuracy. Specifically, the more saturated the machine electromagnetic torque Te , and main flux λm are presented in
is, the larger the numerical error should be. To verify that IM1 Fig. 6. A magnified view of the first peak of ias after the clear-
is well saturated, the dynamic values of λm (see Fig. 3) are ing of the upstream fault is also shown in Fig. 7. Despite the
depicted on the steady-state saturation curve in Fig. 5. The air- unfavorable machine parameters, all models except qd2 again
gap line is also shown in Fig. 5 to better visualize the level of predict electrical and mechanical variables with sufficient nu-
saturation at different operating points. Analyzing Figs. 3 and merical accuracy.
5, it is seen that IM1 is well saturated in steady state (λm = The dynamic values of λm during the study (see Fig. 7) are
1.77 Wb), and that it saturates even further (λm = 1.83 Wb) depicted on the steady-state saturation curve in Fig. 8 along with
shortly after the fault. The machine remains saturated during the air-gap line. The machine IM2 is also well saturated in steady
the rest of the sag. state (λm = 0.473 Wb). Higher values of flux are reached at the
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 9

TABLE III
CALCULATED 2-NORM RELATIVE ERRORS OF ia s , ω r , T e , AND λm
FOR THE SINGLE-PHASE VOLTAGE SAG (IM2)

Model VP- CP-VBR- CP-VBR- qd1 qd2


VBR BDF1 BDF2 (R s n = (R s n =
300Ω) 3Ω)
Var.

i a s (%) 0.069 0.754 0.094 0.473 3.64


ω r (%) 0.002 0.009 0.002 0.017 0.371
T e (%) 0.074 0.273 0.086 0.256 4.406
λm (%) 0.013 0.068 0.014 0.086 3.694

beginning and the end of the voltage sag (λm = 0.488 and 0.484
Wb, respectively). The motor also remains saturated throughout
the rest of the sag.
The 2-norm relative errors of ias , ωr , Te , and λm with respect
to the reference trajectories are summarized in Table III. All
models except qd2 provide accurate results with errors smaller
than 1%. However, there is now a noticeable difference between
the two CP-VBR formulations. Specifically, CP-VBR-BDF2
yields more accurate results than CP-VBR-BDF1 (e.g., 0.094%
error in ias compared to 0.754%) due to its second-order approx-
imation of piq ds . Nevertheless, the numerical accuracy attained
by the CP-VBR-BDF1 model remains satisfactory for most
Fig. 6. Transients in stator current ia s , rotor electrical speed ω r , electromag- practical cases. This study demonstrates that highly accurate
netic torque T e , and main flux λm during a single-phase voltage sag (IM2).
results can be achieved with the proposed constant-parameter
model even for well saturated machines with unfavorable pa-
rameters.

C. Model Accuracy versus Integration Step Size


It is also useful to validate the accuracy of the proposed
models as a function of a constant step size Δt. For this purpose,
the same voltage sag transient study is repeated here using the
proposed CP-VBR-BDF1 and CP-VBR-BDF2 models. To focus
only on the machine models, the whole ac network (Thévenin
equivalent, transformer, and cable) is replaced by an infinite
bus. A third machine (IM3), whose value of Lm u /LΣlr is in
the middle range (43.3), is also considered. The parameters
Fig. 7. Magnified view of the first peak of stator current ia s following the of IM3 are summarized in Appendix C. To enforce a fixed
clearing of the single-phase fault (IM2).
step size using PLECS [1], all simulations were executed using
Simulink’s ode45 solver with relaxed error tolerances.
The resulting 2-norm relative errors of ias using CP-
VBR-BDF1 and CP-VBR-BDF2 are plotted in Fig. 9 for
the three machines. As it can be observed, the errors in-
crease with the step size, which was predicted in Section
V-A. The errors also converge towards zero for small step
sizes, which demonstrates the consistency [31] of the pro-
posed method. At the same time, the errors grow faster for
machines with lower ratios Lm u /LΣlr . This emphasizes the
correlation between model accuracy and machine parameters
noted in Section V-B and observed in Sections VI-A and
VI-B. It can also be seen in Fig. 9 that as the error in-
Fig. 8. Saturation curve of IM2 and its operating range (denoted by the thick creases, CP-VBR-BDF2 is clearly more accurate than CP-VBR-
line) during a single-phase voltage sag. BDF1. For example, for the machine IM2 and Δt = 0.5 ms,
the CP-VBR-BDF2 model yields an error twice as small as the
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE TRANSACTIONS ON ENERGY CONVERSION

overhead. Additionally, the computational cost of the qd model


per time step is much less than that of CP-VBR (approximately
172 μs versus 317 μs for the small system with IM1), which is
due in part to its efficient implementation of saturation [6]. How-
ever, to obtain an accurate solution, the large snubber required to
interface the qd model considerably increases the system’s nu-
merical stiffness, which in turn results in the use of smaller step
sizes and increased computational burden. The proposed CP-
VBR model (in particular the CP-VBR-BDF2 implementation)
therefore offers the best combination of numerical accuracy and
efficiency.

VII. DISCUSSION
A. Classical General-Purpose qd Model
Equations (1)–(18) summarize the traditional/classical
Fig. 9. Cumulative 2-norm relative error of stator current ia s for constant step general-purpose lumped-parameter symmetrical induction ma-
sizes using: (a) CP-VBR-BDF1 and (b) CP-VBR-BDF2. chine model in qd-coordinates. This well-established model
(or its variations) has (have) been extensively used for
TABLE IV several decades to evaluate the dynamics of systems contain-
NUMERICAL EFFICIENCY OF THE CONSIDERED MODELS ing induction machines [5], [15], [35]. It has been experi-
FOR THE SINGLE-PHASE VOLTAGE SAG STUDY mentally validated in numerous publications for various ma-
chines and transient phenomena [8], [10], [35], [36]. To this
VP- CP-VBR- CP-VBR- qd1 qd2 day, the general-purpose lumped-parameter symmetrical in-
VBR BDF1 BDF2
duction machine model remains the default model in many
IM1 No. time 603 567 578 2528 741 commercial-grade electromagnetic transient simulation pro-
steps
CPU time 836 179 184 422 131
grams such as PLECS [1], SimPowerSystems [2], EMTP-RV
(ms) [3], and PSCAD/EMTDC [4], which are widely used by engi-
CPU time 1386 316 318 167 177 neers and researchers in industry and academia.
per time step
(μs) At the same time, there are many higher-fidelity and/or
IM2 No. time 777 816 793 3602 1308 higher-order models capable of reproducing the dynamics of
steps
CPU time 1079 184 197 433 158 physical induction machines more accurately [7], [9], [13],
(ms) [33], [37]–[40]. In some cases, the general-purpose lumped-
CPU time 1389 225 248 120 121
per time step
parameter model is modified as to take into account specific
(μs) phenomena, e.g., main flux saturation harmonics [33], [37], [38]
and leakage inductance saturation [7], [9], [13]. Other models
are based on the finite element [39] or magnetic equivalent cir-
CP-VBR-BDF1 model (0.46% and 0.92%, respectively). An- cuit [40] approaches, which typically allow for a finer represen-
other important point is that all errors remain relatively small tation of the physical structure of machines. Such higher-fidelity
(below 2%) even at a fairly large integration step size of one and/or higher-order models generally require more parameters
millisecond with machine IM2 having unfavorable parameters and corresponding sophisticated procedures for their determi-
(i.e., low Lm u /LΣlr ). In particular, in this study, the error never nation, and therefore are typically used only when the appro-
exceeds 1% for the CP-VBR-BDF2 model. priate high fidelity is necessary. Such models also increase the
computational cost of the solution and may not be justified for
D. Computational Performance system-level studies.

To evaluate the models’ computational performance, the re-


sulting number of time steps, CPU time, and average CPU time B. Impact of Main Flux Saturation
per time step from the studies of Sections VI-A and VI-B are Including main flux saturation in induction machine mod-
summarized in Table IV. For a given machine, all three VBR els is very common. The unsaturated models tend to underes-
models take similar numbers of time steps. However, the CPU timate steady-state stator currents, especially under low load
time of VP-VBR is much higher than that of CP-VBR. For exam- conditions. To demonstrate this point, the voltage sag study of
ple, for IM1, the VP-VBR model takes more than 4 times longer Section VI-B is conducted again with a low load (Tm = 0.2 p.u.)
than CP-VBR-BDF1 or CP-VBR-BDF2 (836 ms versus 179 ms using saturated and unsaturated models, and sufficiently small
or 184 ms, respectively). This is explained by the presence of integration step sizes as to ensure negligible numerical error.
variable inductances in VP-VBR. Table IV also illustrates that The resulting stator current ias is presented in Fig. 10, wherein
using BDF2 instead of BDF1 to approximate piq ds adds little it is seen that in steady state the unsaturated model predicts a
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 11

users do not need to provide new input data. Furthermore, unlike


the qd model, the CP-VBR model does not require interfacing
snubbers, whose parameters would have to be selected as a
compromise between numerical accuracy and efficiency. The
constant-parameter decoupled interfacing circuit of the new sat-
urable model is not only beneficial from a computational cost
point of view (as evidenced by Table IV), it is also practical since
the components library of some simulation programs (such as
SimPowerSystems [2]) do not include time-dependent induc-
tances. Moreover, as demonstrated in Section VI, the gains in
numerical efficiency of the proposed model come at the expense
of very little numerical error.
Fig. 10. Transients in stator current ia s for the study of Section VI-A under It is envisioned that the proposed saturable model could re-
low-load condition as predicted by unsaturated and saturated induction machine
models. place or complement the existing qd and VP-VBR models cur-
rently available in commonly used state-variable-based simu-
current roughly 30% smaller than the saturated model. There lation programs such as SimPowerSystems and PLECS, with
is also a noticeable difference in terms of phase and magnitude essentially no structural changes to the programs. The resulting
between the transient currents predicted by both models after gains in simulation speed can have a significant impact for many
the voltage sag. Another simple and effective mean to observe industrial and research users, in particular when simulating large
the effect of main flux saturation on machine variables is to ap- networks and/or running batch simulations, e.g., Monte Carlo
ply a voltage step which brings the machine into the saturated analyses and parametric sweeps.
region. One such study is presented in [41, Figs. 6–9], in which
significant differences between the saturated and unsaturated VIII. CONCLUSION
stator currents and main fluxes are observed. The study in [41] A new explicit constant-parameter VBR induction machine
also underlines the much more limited effect of saturation on model (CP-VBR) incorporating main flux saturation has been
the rotor speed and electromagnetic torque of grid-connected presented. The model is obtained by transferring the saturation-
machines. dependent (time-varying) part of the equivalent stator induc-
More specifically, representation of main flux saturation is tance into the subtransient voltage source and approximating its
essential when simulating self-excited induction generators [8], corresponding derivative using backward differentiation formu-
[10]. [36]. Its impact on the study of reclosing transients of las (BDFs). The numerical error introduced by this approxima-
induction machines with power factor correction shunt capac- tion is analyzed. It is shown that the error depends on machine
itors has also been recognized for a long time [35]. Addi- parameters and saturation level, and decreases very rapidly with
tionally, some motor drive schemes take main flux saturation the integration step size. Simulation studies of machines with a
into account [5], [11]. For example, it is shown in [5, Section diverse range of parameters are presented. It is shown that even
14.7] that an erroneous estimation of Lm (e.g., arising from a for machines with unfavorable parameters (low Lm u /LΣlr ra-
change in the level of saturation) yields degraded transient per- tio), the introduced cumulative error remains very small (under
formance with indirect rotor field-oriented control approaches. 1 to 2%) for step sizes as large as 1 ms. The proposed CP-
Machine models with main flux saturation are thus required VBR model is more efficient than the state-of-the-art models
when simulating/testing these schemes. For these reasons, the since it does not increase the system’s stiffness (as does the
built-in induction machine models found in the state-of-the- snubber-interfaced classical qd models) and does not require to
art electromagnetic transient programs [1]–[4] all have provi- change/update the state matrices (as does the algebraically exact
sions to incorporate saturation of the main flux’s fundamental variable-parameter VBR model). Overall, the new model offers
component. an excellent combination of numerical accuracy and efficiency,
The main disadvantage of representing main flux saturation while using a constant-parameter interfacing circuit which is
is the additional computational cost incurred. To quantify this easy to use in most state-variable-based transient simulation
overhead, the study of Section VI-B is repeated using the CP- programs. It is envisioned that the proposed model will find
VBR-BDF2 model and the standard unsaturated VBR model its application in many commercial-grade transient simulation
[19]. It is seen that for this scenario the CPU time per time step programs used by thousands of users.
increases by roughly 60% when representing saturation. While
the added computational burden is not negligible, it is believed
APPENDIX A
to be a fair compromise to improve model accuracy.
INDUCTION MACHINE IM1 [10]
1) Machine parameters (converted using [27]): 7.5 kW,
C. Benefits and Potential Use of the Proposed Model 50 Hz, 2 poles, wye-connected, Vlls = 658 V, rs =
The proposed CP-VBR model requires the same typical ma- 1.969 Ω, Lls = 10.2 mH, Lm = 598 mH, rr 1 = 1.093 Ω,
chine parameters as the traditional lumped-parameter qd mod- Llr 1 = 8.98 mH, rr 2 = 5.727 Ω, Llr 2 = 4.07 mH, and
els currently available in commercial programs. Consequently, J = 0.04 J · s2 .
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE TRANSACTIONS ON ENERGY CONVERSION

2) Saturation Characteristic (peak values): [8] E. Levi, “A unified approach to main flux saturation modeling in d-q axis
models of induction machines,” IEEE Trans. Energy Convers., vol. 10,
λm (Wb) 1.237 1.579 1.674 1.754 1.812 1.851 1.881 1.903 1.924 no. 3, pp. 455–461, Sep. 1995.
i m (A) 2.121 3.182 3.677 4.243 4.808 5.303 5.798 6.222 6.718 [9] T. A. Lipo and A. Consoli, “Modeling and simulation of induction motors
with saturable leakage reactances,” IEEE Trans. Ind. Appl., vol. IA-20,
no. 1, pp. 180–189, Jan./Feb. 1984.
3) Transformer: SB = 50 kVA, VH = 12.47 kV, VL = 658 [10] E. Levi, “Main flux saturation modelling in double-cage and deep-
bar induction machines,” IEEE Trans. Energy Convers., vol. 11, no. 2,
V, Z0 = Z1 = 1.5 + 2j%. pp. 305–311, Jun. 1996.
4) Cable: Z0 = 0.15 + j0.15 Ω, Z1 = 0.05 + j0.05 Ω. [11] E. Levi and M. Wang, “Online identification of the mutual inductance for
5) Thévenin equivalent: Vnom = 12.47 kV, Z0 = 0.75 + vector controlled induction motor drives,” IEEE Trans. Energy Convers.,
vol. 18, no. 2, pp. 299–305, Jun. 2003.
j1.95 Ω, Z1 = 0.3 + 0.6j Ω. [12] L. Monjo, F. Córcoles, and J. Pedra, “Saturation effects on torque- and
current-slip curves of squirrel-cage induction motors,” IEEE Trans. Energy
APPENDIX B Convers., vol. 28, no. 1, pp. 243–254, Mar. 2013.
[13] G. J. Rogers and D. Shirmohammadi, “Induction machine modelling for
INDUCTION MACHINE IM2 [9] electromagnetic transient program,” IEEE Trans. Energy Convers., vol. 2,
no. 4, pp. 622–628, Dec. 1987.
1) Machine parameters: 5 hp, 60 Hz, 4 poles, wye- [14] S. D. Sudhoff, D. C. Aliprantis, B. T. Kuhn, and P. L. Chapman, “Ex-
connected, Vlls = 230 V, rs = 0.4122 Ω, Xls = 1.1 Ω, perimental characterization procedure for use with an advanced induction
Xm = 15.7 Ω, rr = 0.4976 Ω, Xlr = 1.1 Ω, and J = machine model,” IEEE Trans. Energy Convers., vol. 18, no. 1, pp. 48–56,
Mar. 2003.
0.11 J · s2 . [15] P. C. Krause and C. H. Thomas, “Simulation of symmetrical induction
2) Saturation characteristic (peak values): machinery,” IEEE Trans. Power App. Syst., vol. PAS-84, no. 11, pp. 1038–
1053, Nov. 1965.
λm (Wb) 0.147 0.295 0.398 0.454 0.486 0.522 0.535 0.543 0.553 [16] L. Wang, J. Jatskevich, V. Dinavahi, H. W. Dommel, J. A. Martinez, K.
i m (A) 3.536 7.071 10.61 14.41 17.68 24.75 28.28 31.82 35.82 Strunz, M. Rioual, G. W. Chang, and R. Iravani, “Methods of interfacing
rotating machine models in transient simulation programs,” IEEE Trans.
Power Del., vol. 25, no. 2, pp. 891–903, Apr. 2010.
3) Transformer: SB = 25 kVA, VH = 12.47 kV, VL = 230 [17] O. Wasynczuk and S. D. Sudhoff, “Automated state model generation
V, and Z0 = Z1 = 1.5 + j2%. algorithm for power circuits and systems,” IEEE Trans. Power Syst.,
vol. 11, no. 4, pp. 1951–1956, Nov. 1996.
4) Cable: Z0 = 0.15 + j0.15 Ω, Z1 = 0.05 + j0.05 Ω. [18] S. D. Pekarek, O. Wasynczuk, and H. J. Hegner, “An efficient and accurate
5) Thévenin equivalent: Vnom = 12.47 kV, Z0 = 0.75 + model for the simulation and analysis of synchronous machine/converter
j1.95 Ω, Z1 = 0.3 + j0.6 Ω. systems,” IEEE Trans. Energy Convers., vol. 13, no. 1, pp. 42–48,
Mar. 1998.
[19] L. Wang, J. Jatskevich, and S. D. Pekarek, “Modeling of induction ma-
APPENDIX C chines using a voltage-behind-reactance formulation,” IEEE Trans. Energy
INDUCTION MACHINE IM3 [5] Convers., vol. 23, no. 2, pp. 382–392, Jun. 2008.
[20] S. D. Pekarek and E. A. Walters, “An accurate method of neglecting
1) Machine parameters: 50 hp, 60 Hz, 4 poles, wye- dynamic saliency of synchronous machines in power electronic based
systems,” IEEE Trans. Energy Convers., vol. 14, no. 4, pp. 1177–1183,
connected, Vlls = 460 V, rs = 0.087 Ω, Xls = 0.302 Dec. 1999.
Ω, Xm = 13.08 Ω, rr = 0.228 Ω, Xlr = 0.302 Ω, and [21] S. D. Pekarek, M. T. Lemanski, and E. A. Walters, “On the use of
J = 1.662 J·s2 . singular perturbations to neglect the dynamic saliency of synchronous
machines,” IEEE Trans. Energy Convers., vol. 17, no. 3, pp. 385–391,
2) Saturation characteristic (peak values): Sep. 2002.
[22] M. Chapariha, F. Therrien, J. Jatskevich, and H. W. Dommel, “Explicit
λm (Wb) 0.38 0.48 0.60 0.68 0.75 0.79 0.82 0.86 0.92 formulations for constant-parameter voltage-behind-reactance interfacing
i m (A) 11.20 14.33 18.33 21.31 24.42 26.74 29.07 33.21 40.78 of synchronous machine models,” IEEE Trans. Energy Convers., vol. 28,
λm (Wb) 0.96 1.01 1.06 1.10 / no. 4, pp. 1053–1063, Dec. 2013.
i m (A) 46.21 53.21 60.33 66.10 / [23] F. Therrien, M. Chapariha, and J. Jatskevich, “Generalized state-space
saturable induction machine model using a voltage-behind-reactance for-
mulation,” in Proc. IEEE Power Energy Soc. Gen. Meeting, Vancouver,
BC, Canada, Jul. 21–25, 2013, pp. 1–5.
REFERENCES [24] D. C. Aliprantis, O. Wasynczuk, and C. D. Rodrı́guez Valdez, “A voltage-
behind-reactance synchronous machine model with saturation and arbi-
[1] Plexim GmbH. (2013). Piece-wise linear electrical circuit simulation trary rotor network representation,” IEEE Trans. Energy Convers., vol. 23,
(PLECS) user manual version 3.4, Plexim GmbH, Zurich, Switzerland no. 2, pp. 499–508, Jun. 2008.
[Online]. Available: www.plexim.com [25] M. Chapariha, L. Wang, J. Jatskevich, H. W. Dommel, and S. D.
[2] SimPowerSystems R2009a—User’s Guide, Hydro-Québec and The Math- Pekarek, “Constant-parameter RL-branch equivalent circuit for inter-
Works, Inc., Natick, MA, USA, 2009. facing AC machine models in state-variable-based simulation pack-
[3] CEA Technologies, Inc. (2007). Electromagnetic Transient Program, ages,” IEEE Trans. Energy Convers., vol. 27, no. 3, pp. 634–645,
EMTP-RV [Online]. Available: www.emtp.com Sep. 2012.
[4] Manitoba HVDC Research Centre and RTDS Technologies, Inc. (2009). [26] S. J. Chapman, Electric Machinery Fundamentals, 4th ed. New York, NY,
PSCAD/EMTDC V4.2 On-Line Help [Online]. Available: www.pscad.com USA: McGraw-Hill, 2005.
[5] P. C. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of Electric [27] J. Pedra, I. Candela, and L. Sainz, “Modelling of squirrel-cage induction
Machinery and Drive Systems, 2nd ed. Piscataway, NJ, USA: IEEE Press, motors for electromagnetic transient programs,” IET Elect. Power Appl.,
2002. vol. 3, no. 2, pp. 111–122, Mar. 2009.
[6] F. Therrien, L. Wang, J. Jatskevich, and O. Wasynczuk, “Efficient explicit [28] S. D. Pekarek, E. A. Walters, and B. T. Kuhn, “An efficient and accurate
representation of AC machines main flux saturation in state-variable-based method of representing magnetic saturation in physical-variable models
transient simulation packages,” IEEE Trans. Energy Convers., vol. 28, of synchronous machines,” IEEE Trans. Energy Convers., vol. 14, no. 1,
no. 2, pp. 380–393, Jun. 2013. pp. 72–79, Mar. 1999.
[7] S. D. Sudhoff, D. C. Aliprantis, B. T. Kuhn, and P. L. Chapman, “An [29] Simulink 7—User’s Guide, The MathWorks Inc., Natick, MA, USA,
induction machine model for predicting inverter-machine interaction,” 2009.
IEEE Trans. Energy Convers., vol. 17, no. 2, pp. 203–210, Jun. 2002.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

THERRIEN et al.: CONSTANT-PARAMETER VBR INDUCTION MACHINE MODEL INCLUDING MAIN FLUX SATURATION 13

[30] M. Chapariha, F. Therrien, J. Jatskevich, and H. W. Dommel, “Imple- Mehrdad Chapariha (S’08) received the B.Sc. and
mentation of induction machine VBR model with optional zero-sequence M.Sc. degrees in electrical engineering from the Is-
in SimPowerSystems, ASMG, and PLECS toolboxes,” in Proc. Int. fahan University of Technology, Isfahan, Iran, in
Conf. Power Syst. Transients, Vancouver, BC, Canada, Jul. 18–20, 2013, 2006 and 2009, respectively, and the Ph.D. degree in
pp. 1–5. electrical engineering from the University of British
[31] U. M. Ascher and L. R. Petzold, Computer Methods for Ordinary Differ- Columbia, Vancouver, BC, Canada, in 2013.
ential Equations and Differential-Algebraic Equations. Philadelphia, PA, He is currently a Postdoctoral Fellow and a Ses-
USA: SIAM, 1998. sional Lecturer at the University of British Columbia.
[32] E. Levi, “Impact of cross-saturation on accuracy of saturated induction His current research interests include modeling elec-
machine models,” IEEE Trans. Energy Convers., vol. 12, no. 3, pp. 211– trical machinery, simulation of power systems tran-
216, Sep. 1997. sients, modeling and design of electrical drive and
[33] X. Tu, L.-A. Dessaint, R. Champagne, and K. Al-Haddad, “Transient mod- power electronic systems, and applications of power electronics in power sys-
eling of squirrel-cage induction machine considering air-gap flux satura- tems.
tion harmonics,” IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 2798–2809,
Jul. 2008.
[34] U. M. Ascher and C. Greif, A First Course in Numerical Methods. Philadel-
phia, PA, USA: SIAM, 2011.
[35] F. P. De Mello and G. W. Walsh, “Reclosing transients in induction motors
with terminal capacitors,” AIEE Trans. Power App. Syst., vol. 80, pp. 1206–
1213, Feb. 1961.
[36] K.-E. Hallenius, P. Vas, and J. E. Brown, “The analysis of saturated self-
excited asynchronous generator,” IEEE. Trans. Energy Convers., vol. 6,
no. 2, pp. 336–345, Jun. 1991.
[37] J. C. Moreira and T. A. Lipo, “Modeling of saturated ac machines includ-
ing air-gap flux harmonic components,” IEEE Trans. Ind. Appl., vol. 28,
no. 2, pp. 343–349, Mar./Apr. 1992.
[38] D. Bispo, L. M. Neto, J. T. de Resende, and D. A. de Andrade, “A Juri Jatskevich (M’99–SM’07) received the
new strategy for induction machine modeling taking into account the M.S.E.E. and the Ph.D. degrees in electrical engi-
magnetic saturation,” IEEE Trans. Ind. Appl., vol. 37, no. 6, pp. 1710– neering from Purdue University, West Lafayette, IN,
1719, Nov./Dec. 2001. USA, in 1997 and 1999, respectively.
[39] S. J. Salon, Finite Element Analysis of Electrical Machines. Norwell, MA, Since 2002, he has been a Faculty Member with
USA: Kluwer, 1995. the University of British Columbia, Vancouver, BC,
[40] S. D. Sudhoff, B. T. Kuhn, K. A. Corzine, and B. T. Branecky, “Magnetic Canada, where he is currently a Professor of Electri-
equivalent circuit modeling of induction motors,” IEEE Trans. Energy cal and Computer Engineering. His current research
Convers., vol. 22, no. 2, pp. 259–270, Jun. 2007. interests include power electronic systems, electrical
[41] L. Wang and J. Jatskevich, “Including magnetic saturation in voltage- machines and drives, and modeling and simulation of
behind-reactance induction machine model for EMTP-type solution,” electromagnetic transients.
IEEE Trans. Power Syst., vol. 25, no. 2, pp. 975–987, May 2010. Dr. Jatskevich chaired the IEEE CAS Power Systems and Power Electronic
Circuits Technical Committee in 2009–2010. He was as an Associate Editor
for the IEEE TRANSACTIONS ON POWER ELECTRONICS from 2008 to 2013,
and is currently the Editor-in-Chief of the IEEE TRANSACTIONS ON ENERGY
CONVERSION and the Editor of the IEEE POWER ENGINEERING LETTERS. He is
Francis Therrien (S’12) received the B.Eng. degree the Chair of the IEEE Task Force on Dynamic Average Modeling, under the
in electrical engineering from the Université de Sher- Working Group on Modeling and Analysis of System Transients Using Digital
brooke, Sherbrooke, QC, Canada, in 2010. He is cur- Programs.
rently working toward the Ph.D. degree in electrical
and computer engineering at the University of British
Columbia, Vancouver, BC, Canada.
He was a Power Systems Researcher for CYME
International T&D until 2011. His current research
interests include modeling of electrical distribution
systems, electrical machines, and simulation of elec-
tromagnetic transients.

You might also like