Você está na página 1de 10

IEEE Transactions on Power Apparatus and Systems, Vol. PAS-101, No.

6 June 1982

1 342

UNIVERSAL MACHINE MODELING FOR THE REPRESENTATION OF


ELECTRIC MACHINERY IN AN ELECTROMAGNETIC TRANSIENTS
Hian K. Lauw, Member, IEEE
Dept. of Electrical & Computer Engineering
Oregon State University
Corvallis, Oregon 97330

ABSTRACT - This paper describes the theory, development


and practical application of a new universal machine
module, which can be used for the transient analysis of
motor/generator phenomena in large and complex power
networks. In the general case, doubly-fed electric
machines with an arbitrary number of coils and magnetic
saturation can be represented. Included are all of the
major classes which are of common industrial interest:
synchronous and induction machines of 1, 2, or 3
phases, and dc machines. A procedure for maximizing
the flexibility of the machine-network interface is
explained.
Computer implementation has been variablydimensioned, so that any number of motors or generators
can be interconnected with mechanical systems and
electric networks of arbitrary size and configuration.
Universal machine modeling is now a standard feature of
the BPA Electromagnetic Transients Program (EMTP),
which is widely used on numerous different computers.
SYMBOL NOTATION
Index "UM" denotes "universal machine"
Index "d" denotes a direct-axis quantity
Index "q" denotes a quadrature-axis quantity
Index "P" denotes a power-coil quantity
Index "E" denotes an excitation-coil quantity
Square brackets indicate square matrices (e.g.,
Underlined symbols are vectors
V
= voltage
= flux linkage
= current
i
= resistance
R
= leakage inductance
Lt
Lm, Xm = main inductance and main flux linkage
itk] = k - dimensional identity matrix
= number of pole pairs
np
= angular position of the rotor
= angular speed of the rotor

"FAI")

1. INTRODUCTION

There has been increased interest in recent years


to include several major types of rotating electrical
machines in general simulation programs for digital
Such modeling would permit detailed study
computers.
of transients involving the interaction of electric
machines with an electric power grid, as well as the
machine interaction with complex networks of mechanical
Interest is particularly motivated by the
components.
demand for feasibility studies which include detailed
representation for the interface and control variables.
Such detail is required for the design of reliable and
economical energy conversion based on erratic energy
resources such as wind, hydro, active solar, and the

ROTATING
PROGRAM

W. Scott Meyer, Member, IEEE


Bonneville Power Administration, EOGB
P. 0. Box 3621
Portland, Oregon 97208
breaking energy of electric cars.
paper responds to this need.

Work of the present

If separate computer coding were to be written for


each of the different types of electric machines, the
result would have been prohibitively cumbersome, and
wasteful of memory. This is particularly true when one
considers the different required interfaces between the
machines and the electric power grid, and between the
machines and associated mechanical systems. For these
reasons, a unified, generalized approach was instead
pursued, leading to a single universal machine (UM)
component or module, and a single standard interface
for connection with the electrical and the mechanical
networks. Thus a single component has been designed to
serve all machinery modeling needs of the industry.
A binary set of control parameters was chosen to
specify which of the possible features of the UM might
be desired by any particular user, for any particular
representation. Available choices range from a general
doubly-fed electromechanical energy converter of quite
arbitrary specification to the common major classes of
rotating electric machines: synchronous machines (SM),
induction machines(IM), and direct current machines
(DM). For the SM or IM choice, either one-, two-, or
three-phase types are allowed; in the DM case, various
types of usage are covered, such as separately-excited,
series, shunt, and compound field connections.

Early in 1980, UM modeling was added to the BPA


Electromagnetic Transients Program (EMTP), which is
used on numerous computer systems for the simulation of
transients by the electric utility industry r141.
Because this digital computer program has no fixed
table sizes, the user can simulate an integrated system
involving any number and type of electric machines, any
configuration and size of electric network or connected
mechanical system, and a variety of special components
of the power grid F131. Of course, computer running
time and memory requirements increase with the size and
sophistication of the modeling.
Typical applications of the new UM modeling which
most important to power utilities r,31 include the
subsynchronous resonance problem of series-capacitor
compensation, studies of transmission line reclosure,
independent-pole switching, high-speed automatic line
reclosure, energization, synchronizing, load rejection
or unit tripping, the loss of synchronism, unbalanced
system conditions, the stalling of cooling pump drives,
etc. Moreover, the past decade has seen increased use
of power electronic components to control electric
machinery and to condition the output energy. Detailed
studies of the interaction of various types of machines
with an arbitrary power electronic converter and its
firing control strategy can be conveniently performed
using the EMTP.
are

This paper summarizes UM development and usage


within the EMTP framework.
Included are a detailed
A paper recommended and approved by the
81 SM 430-8
formulation of the mathematical model, the selection of
IEEE Power System Engineering Committee of the IEEE
particular special cases (SM, IM, or DM), the numerical
Power Engineering Society for presentation at the
solution scheme, and considerations of interface with
IEEE PES Summer Meeting, Portland, Oregon, July 26-31,
the electric power grid and the associated mechanical
1981. Manuscript submitted February 3, 1981; made
system. A practical industrial study involving wind
available for printing June 9, 1981.
generation is summarized as illustration of the theory.
0018-9510/82/0600-1342$00.75 1982 IEEE

1343

2. THE

UNIVERSAL MACHINE

MODEL

Although electric machines come in various sizes


and types, it has long been recognized that they all
operate on the same fundamental principle. This is the
tendency of two magnets to align themselves, which led
to the development of "generalized machine theory" as
comprehensively documented in Ref. r3]. Although the
underlying assumptions impose some limitations (e.g.,
lunpedl-po-lr,r.mete?r :rod4liLng), Ihis tv-h ory ha. proven to
be most useful for the representation of synchronous,
induction, and dc machines in system studies of the
industry. The UM is based on this theory, as shall
now be documented.
UM theory rests on the foundation of a generalized
machine model with equivalent electrical coils for the
direct, the quadrature, and the zero-sequence axes
(i.e., Park's variables r41 are used). The following
specific assumptions are applicable:
(1) The-power coils are confined to the rotor (for
purposes of mathematical convenience only), and are
limited to a maximum of three. The UM represents these
using three stationary coils (call them the D-coil, the
Q-coil, and the OP-coil) which are connected to three
pairs of brushes belonging to ideal commutators.
(2) The stator is allowed to have just one single
zero-sequence coil OE,plus an arbitrary number of coils
on the D-axis and the Q-axis. These excitation coils
shall be denoted as the OP-coil and
. . . . . .
D -coil
Dl1-coil, D
Ql-coil, Q2-coil, . . . . Q -coilh
(3) The OP-coil and the OE-coil are called the
zero-sequence power and excitation coils, respectively.
They are not magnetically coupled with each other, or
with any other coils of the model. All D-coils share
the direct axis as a common magnetic circuit, while all
Q-coils share the quadrature axis.
(4) The configuration and size of the mechanical
system are arbitrary, since only a net interface torque
is required by the UM equations.
Mechanical dynamics
are represented elsewhere (see Section 3).
(5) Saliency is confined to the excitation side.

It is worth emphasizing the significance of these


UM modeling, assumptions. Nothing has been said about
connections of the coil terminals, which can be quite
arbitrary (see Section 3). The limitation to not more
than three power coils (point 1) is obviously justified
by typical present-day practice. On the other hand,
the open-ended excitation modeling (point 2) is crucial
to UM flexibility, and it goes far beyond the simple
observation that different machines possess different
numbers of excitation coils.
For transient studies
involving unbalanced loads, or faults, or energization,
eddy current losses are known to be important in the
excitation portion of the machine. An array of shortcircuited excitation coils can be used to model such
phenomena F91. The lack of any mechanical restrictions
(Point 4) has particular significance for resonance
studies which involve interaction of the mechanical
system with the connected electric power grid. Any one
assumed set of standard assumptions about mechanical
modeling would almost certainly not satisfy all of the
subsynchronous resonance (SSR) research needs of the
electric power industry rf131. Finally, the prohibition
against saliency in the power circuit is reasonable
physically (power coils are usually placed on the round
side of a machine), and-it has the effect of greatly
simplifying the mathematical description.
A complete set of differential equations which
reflect these UM assumptions is contained at the start
of Appendix I. Computer implementation in the EMTP is
based on these, as well as certain added assumptions of
numerical analysis. The trapezoidal rule was selected
for integration because of proven numerical stability,

and its documented success with the "stiff" electrical


and control systems to which the UM is often connected
ri, 131. Iteration is involved, in order to eliminate
the complicating nonlinear effects of rotor speed. Due
to the use of main and leakage inductances rather than
self and mutual ones, the resulting recursive equations
(A5)-(A6) involve the inversion of only 3x3 matrices at
each time step. Each set of UM equations can be solved
as an isolated nonlinear system at each time step, due
to the decoupling effect of all distributed-parameter
transmission line modeling Fl, 13].
3.

THE

UM

INTERFACE

Generally the UM is an electromechanical energyconversion device which is always connected to an


electrical network and a mechanical system. A schematic diagram of the system to structure the discussion in
this section is depicted in Fig.1 As shown in the figure, the system is decomposed into four modules: the
Electric Network (EN), the Mechanical System (MS), the
Control System (TACS) and the UM.
The EN and TACS modules are existing features of
the EMTP which has been reported elsewhere H1,51. They
include various power system components with their own
such
as
transformers,
characteristics'
peculiar
transmission lines with lumped as well as distributed,
frequency dependent, line parameters, cables, surge arresters, circuit breakers, valves for any converter
The numerical integration routine is
configuration.
the trapezoidal rule, which, due to its property of
numerical stability, is proven to be adequate to solve
Moreover, the
the power system's stiff equations.
use of optimal triangular factorization to exploit the
system's sparsity, and implementation of 'variable dimensioning, result in high computational efficiency.
As mentioned in the previous section, the MS is
also supposed to include the entire mechanical portion
of the UM; the UM contains the electromagnetic part.
Therefore, the dual variables involved in the decomposition of the MS and UM are the mechanical speed of the
, and the electromagnetic torque
UM rotor mass, w
T
produced by t%me electomagnetic system of the UM.
Since any mechanical system which can be described by
lumped parameters generates a mathematical model of the
same form as electric circuits X61, no special module
needs to be developed for the MS. Immediate use can be
made of the available electric network module after
converting the mechanical system elements into their
equivalent electric network elements. Since this approach realizes the intended generality of the mechanical system of the UM, neither its configuration nor its

MECHAN ICAL

SYSTEM

Wur4; TTUM
I

k-

UNIVERSAL
MACHINE

_-

CONTROL
SYSTEM

TRANSFORMATION

f .Mt
IUMLs
tum
L

ELECTRIC
NETWORK
Fig. 1.

---

---__j

Diagram of UM interface variables.

1344

size is fixed. Moreover, as will be shown in a later


section, no time-step delay is required in interfacing
the mechanical system with the UM.
The interface of the four modules (UM, EN, MS, and
TACS) has two most important functions:
(1). To establish a consistent information flow
between the four modules as well as to realize a valid
composition procedure.
(2). To provide the information to the system as to
which type of the UM is dealt with.
In the next sections these functions of the interface will be discussed in more detail.

4. INFORMATION FLOW AND COMPOSITION


The most obvious information flow is with regard
TACS, the control system. All possible terminal
voltages of the UM are easily made accessable for algebraic supplementation with control signals. In addition all coil currents are made available for signal
processing in the control system.
to

The composition of the UM with the EN and MS modules is however not merely a matter of signal flow.
It involves the synchronization of the dual variables
voltage and current at each electrical terminal seperating the UM from the EN; for the mechanical side the
dual variables are the electromagnetic torque T.,, and
the mechanical speed of the rotor mass Ws1
Therefore,
a choice of the information flow with respect to these
dual variables is required. This choice is of course
completely arbitrary, but due to the solution set up of
the EN in the EMTP, here the information flow is chosen
as indicated in Fig. 1. The EN expects a current injection from the UM, and the terminal voltage is to be
an input to the UM.
In the network representation (analog) of the mechanical system, the torque can be taken to correspond
to current and the mechanical speed to correspond to
voltage. Since, as mentioned earlier, the MS module is
set up exactly similar to the EN module, the information with respect to the mechanical dual variables is
here chosen as indicated in Fig. 1.
In accordance with the chosen information flow,
the decoupled solution of the decomposed system can now
be easily performed. To the UN module, all terminal
voltages and the speed of its rotor mass can be considered as given input driving functions at each instant
of time.
Output variables to be solved by the UM are
the electromagnetic torque and all coil currents.

Although obviously decomposition is realized rather easily, composition of the modules to provide synchronized solutions requires a valid special procedure.
Use is made here of the compensation method r11,71 which
can be applied to decouple and compose nonlinear elements from the linear part of an electic network. This
is basically an application of Thevenin's theorem for
multiports at each discrete time-step of the computation.. The electrical network as well as the mechanical
network are required to contain linear or piecewise linear elements. The UM is the nonlinear set of elements
which should be connected to these linear networks. It
can be remarked that if the electrical and mechanical
networks contain other nonlinear elements, then a reasonable approach is to segregate these elements in a
subnetwork, which seperate them from the linear subnetwork to which the UN is connected. By representing the
connecting lines of the subnetworks with distributed
lines, the subnetworks only interact through history
functions r1l. The procedure for composition of the UM
module with the EN and MS modules will now be shown.

Application of the Trapezoidal Rule gives the following general formulation of the EN:

Fy 1Ev(t)

FY

1 is
if
v
is
i
is
h
is

-E

i(t)

(I)

-E

the model admittance matrix, which is constant


the time step is chosen to be constant.
a column vector of node voltages
a column vector of injected node currents
a column vector of known history functions

Since the MS module is treated in exactly the same way


as the EN module, application of the Trapezoidal Rule
to the differential equations of the mechanical system
will yield the general equation for the MS similar in
form to that of (1):

ry 1
M

w(t)

T(t)

(2)

-M

Thus obviously instead of column vectors for voltages


and currents, we now have column vectors for respectively speeds and torques (or forces).

Now let us denote the voltages and the currents of


the electrical nodes to which the UM is connected with
. They are obviously subrespectively vU and i
vectors of v and i respectively. The composition of
the EN,MS and UM modules in accordance with the desired
information flow as shown in Fig. 1, can now be realFrom the
ized by the compensation method as follows.
EN equations in (1) it is always possible to extract
the relation:
V

M(t)

rzTH(t)1

(t)

(t)

(3)

F ETI
and vTHcan be called respectively the Thevenin
E
impedance -matrix and the Thevenin voltage vector. From
the formulation of (3) it is easily seen that they are
the impedences and voltage sources of the EN as "seen"
from the terminals of the UM, which have the terminal
voltages v .
Furthermore, inspection of (4) immediately showsuEat these Thevenin variables can be solved
form the EN equations in (1) as follows:

vTH(t) = solution of the vujt) from (1) for


i JtY = 0, that is, those entries of i(t) corresponding to the UM currents are to be set to zero.

rZTH(t) = solution of v (t) obtained from (1)


for iuM(t) = 1 and vum(t) = 0, i.e. the right hand
side of (1) are all set to zero except for the entries
of i(t) corresponding to the UM currents, which have to
be set to unity.
Similarly for the MS module we can extract from
the MS equations in (2) the relation analog to (3):

UM(t)
TH

ZT

(t) *

TUM(t) +uTH(t)

(4)

TH

ZM and w can be called the mechanical Thevenin variables.


They are solved in exactly the same way as for
the electrical Thevenin variables, i.e.
respectively
rZ 1 and v . But obviously instead of (1) now the MS
module equations as given by (2) have to be considered.

From the procedure as outlined it follows that all


Thevin variables can be solved at any instant of time t
without any knowledge of the UM variables at this time;
they do of course depend on the history (at time t- At)
of the UM variables. This is crucial to the recursive
solution algorithm which is now to be described.
known
Then:

Suppose that at time t all history functions are


(i.e., all variables at time t - tt are given).

1345
Step 1: Solve as outlines at time t all electrical and mechanical Thevenin variables from (1) and (2)
of the EN and MS modules.

Step 2: The terminal voltages vuM(t) and the


speed w M(t) are now offered to the UM module through
(3) and T4). In combination with the equations of the
UM module given by (A3),(A4),(A5) and (A6), at this moment all output variables of the UM module can be calculated, i.e. all currents ium(t) and the electromagnetic torque TUM(t).
Step (3): Output of the UM module is offered to
the EN and MS modules to solve at time t all unknown
This completes
voltages by respectively (1) and (2).
all required computations at time t and by repeating
Step 1,2 and 3, computations for the next integration
step at time t + At can be pursued.
In step (2) a nonlinear set of UM equations is to
be solved.
Two types of nonlinearities have to be
dealt with: a functional nonlinearity and a product
There are of course several numerical
nonlinearity.
approaches to solve equations containing nonlinear relations between the variables. The saturation nonlinearity appears in (A3), where the main inductances
Lmd and Lm become functions of the magnetizing cur. In the EMTP a piecewise linearizarents i,v mdm
and i
tion approach is taken for this type of nonlinearity.
The second type of nonlinearity, as can be seen from
(Al), is formed by the product of the mechanical speed
WUM(t) and the main fluxes in the d- as well as the
q-axis, i.e. Xn- (t) and X (t). Furthermore, product
terms also come up in (A4) Tor the electromagnetic torque: products of fluxes and currents. This nonlinearity is handled by an iteration loop within the numerical algorithm of the UM module. The iteration loop can
be simply designed as follows:
(a) Predict mechanical speed by linear extrapolation.
(b) solve the UM currents from (A3) and (A5)
(c) find the electromagnetic torque from (A4)
(d) calculate the mechanical speed from (2)
(e) if calculated mechanical speed is not within some
technically acceptable bound of the predicted
Otherwise stop.
value, go back to (a).
Rather than using more sophisticated methods like
Newton-Raphson or Gauss-Seidel, this pragmatic approach
is preferred for algorithmic simplicity and effectiveness considering the parameter-ratio's dealt with by
Due to the relatively
most electric machine usage.
large rotor moment of inertia of practical electric machines, within the usual practically important range of
fluctuations of the electrical variables, the mechanical speed is expected to change only marginally. As a
result, the number of iterations scheme as outlined
will be usually one and not more than three, for most
transient studies of interest r21.

internal UM variables. The required UM input


v uware obtained from the transformed (3):
--UM

(t)

EZTH(t)1
E

where: y (t) = rT(t)1 y(t)


-*
-

FZTH(t)1
rT(t)1

rT(t)1

(t) +

_ UM*

for

VTH(t)

(5)

y = v , i
or vTH
- -UM-UM

rz2 (t)1 rT(t)1

= desired transformation matrix

The real UM output currents which are needed by the


module is simply obtained from:

EN

(6)
iu (t) = ET(t)] i (t)
It is well k-nown from synchroiious machine theory r81
that a power invariant transformation from real 3-phase
variables to o,d,- variables requires the transformation matrix FT(t)l to be:
1 /Y

=/2/3

FT(t)1

1/V
cos(0

sinL

sin(0

1/V
120)

cos(O

120)

120)

sin(o

120)

where 0 = number of pole pairs times the


tion angle

rotor

(7)
posi-

FT(t)1 is orthogonal, i.e. FT(t)it= rT(t)1 with


superscript "t" denoting matrix transpostion. The purpose of transforming the synchronous machines is for
The equivalent o,d, and q
analytical convenience.
power coils are stationary with respect to the excitation coils and decoupling of the zero component from
the d- and q- components is obtained. These d- and qcomponents have magnetic axes perpendicular to each
other. Furthermore, it can be shown that with balanced
three-phase sources, the transformed equations become
time-invariant, simplifying considerably for instance
an analytical stability analysis.
This idea can be generalized to become a powerful
approach in unifying different types of machines and
vice versa, in retrieving these different types from a
unified model.
For this purpose it is to be realized
that the orthogonal transformation matrix of (7) can be
written as the product of 2 likewise orthogonal matrices:

rT(t)l

EP(e)l

(8)

Us]

where:

rp(e)1

o
0

cose

sinO

01

-sine

cose

r SI

i/vT

1/3 //3
-1 //
(9)
_273 -1 /vg
L 0
+1/1fj
-

The rotor position angle e is obviously a


time as the machine is rotating.

5. UM TYPE TRANSFORMATION
In this section the second important function of
the interface will be covered, i.e. to provide the
system with information as to which type of electric
machine is dealt with. The UM module is a generalized
machine model which needs a proper transformation of
its electric variables before it can be used as a model
representing the desired type of machine. This interface function is realized by controlling the information flow between the UM module and the EN module, as
indicated in Fig. 1 by the processor marked with TRANSFORMATION. The information flow concerns the voltages
from the EN to the UM and the currents from the UM to
the EN. Thus in Step 2 of the algorithm as discussed
in the previous section, the input to the UM module is
and the output of the UM module is
v --,rather than v
i UM rather than i um if v UM*and i UMdenote the

voltages

function

Of

Field

pole

(2a)
Fig. 2.

(2b)

Schematic diagram of a synchronous machine.

1346

Now let us inspect the consequence of applying the


of machine variables with the matrices
EP(e)i and rs]. Let us first turn our attention to the
3-phase synchronous machine. The 3-phase power side of
this machine is on the stator and the excitation side
(including damper windings) on the rotor. The stator
periphery is round, whereas the rotor might contain
A schematic diagram is given in
pole saliencies.
Fig. 2a. Conceptionally this configuration is obviously equivalent to that of Fig. 2b, where the 3-phase
power side now is placed on the rotor and the excitaThe 3-phase AC power is
tion side on the stator.
tapped from the rotor from slip rings. The transformation with the [P( 0)1 and the FS1 is applied to the
power coils. Transforming the power coil equations
with the S matrix shows that the 3-phase coils are replaced with 3 equivalent coils: a zero-component coil
decoupled from two other component coils, which have
Subsequent
magnetic axes perpendicular to each other.
transformation with the rP(t)1 matrix is to establish
that these equivalent coils are replaced with equivalent windings which are stationary in space and consequently stationary with respect to the excitation side.
Furthermore, these equivalent windings are chosen such
that their magnetic axes are in line with the d (direct) and q (quadrature) axis, in addition to the o
coil. Thus physically this rP(e)1 transformation is
replacing the slip rings with an ideal commutator. The
stationary equivalent coils, o,d,q, are considered to
represent three coils which are connected to three
The imporpairs of brushes of the ideal commutator.
tant conclusion to be taken from this discussion is the
fact that the transformation with the product of the
[P(e)] and S matrices results in a machine configuration which exactly satifies the UM coil arrangements as
given in Section 2. And vice versa, by back transformation, the power variables of the synchronous machines
can be retrieved from the UM. The excitation-side coil
variables are not transformed at all.

transformation

The approach as outlined can obviously be applied


well to the other types of machines. If for instance 3-phase induction machines are dealt with, then
both the power and the excitation side coils are transformed by the ES] matrix. The subsequent transformation with the FP(e)l matrix is only applied to the
power coil variables . Direct current machines are to
be extracted from the UM without any transformation of
both the power and the excitation side. To have a general interface set-up of the power coils, including
less then 3-phase applications of the UM, only 3 special binary control parameters to the FP() and s
matrices need to be implemented as follows:
as

UM

(t)

(1

EP(OUM)i

c372/3

sineum

0
sineUM

C2

C3 /Y'3

c3//6

-c3 /12

(11)

cosOUM 4

c3 /y/3

.C 3/
Psi

cOseUM

(10)

ci) e(t)

E0
.

-C3/
cl

c2 + c3

rSlt

LP(o

)It

It can be remarked that more control parameters


can be introduced to include for instance several types
of direct current machines such as the series or parallel compound machines. Maximum use of the flexibility
of digital computers is made by simply adding or disregarding required additional terms in the general formulation of the UM equations. The interface as built in
into the EMTP is organized such that the user needs not
By setting a
to bother about the UM configuration.
simple flag to a certain value, a corresponding electric machine type is obtained; the interface selects
the proper combination of the control parameters.
Moreover, the user needs not to bother either about the
fact that the UM set-up places the power-side coils on
the rotor, whereas for certain machine types they
should be located on the stator. To the user, the design configuration of the specific machine type called
for is exactly extracted from the UM by the interface.
This function of the interface is realized along the
lines as discussed in this section.

6. ILLUSTRATIVE

OF WIND

GENERATOR

The characteristics of induction wind generators


which supply electric power to passive loads have been
possible
However, numerous
well documented riol.
schemes to integrate the generator with the mechanical
system, and also with the electric power grid, are
still being actively researched. As an illustration,
we shall consider the system shown schematically in
Energy from the wind generator (on the left)
Fig. 3.
is transferred to the load via a direct current (dc)
link. The synchronous machine (at right) and connected
ac transmission line represent a strong power grid,
which merely serves as a standby power supply for the
Figure 3 also
load, and a synchronizing reference.
shows decomposition of the total system into MS, UM,
EN, and TACS subsystems.
The MS subsystem represents one NASA MOD-2 Wind
Turbine System (WTS), a megawatt-size turbine which is
used by the largest BPA wind generators. There are two
pitch-controlled blades mounted on a hub, a quill
shaft, low-speed and high-speed shafts connected by a
gearbox, and the generator rotor. The representation
of this mechanical subsystem shall be assumed similar
to that of a recent publication r121, resulting in the
Symbols
lumped-parameter electrical analog of Fig. 4.
"B"i, "H", and "G" stand for "blade", "hub", and

Inverter
bridge

Rectifier
bridge

/2

MOD-22
Wind

Turbine

System

(13)

The inverse of the FP(e)1 and Fsl matrices as defined


in (11) and (12) does not exist in the case of 2-phase,

SIMULATION

For several years now, there has been increasing


interest in renewable sources of energy, and hence in
the associated feasibility studies which consider reliability and performance of the conversion to useful
electrical power. Wind energy conversion systems are
particularly popular, and one that is designed around a
capacitor-excited induction generator has been taken to
illustrate EMTP simulation using the new UM modeling.

(12)

C3 is set equal to 1 if a 3-phase ac system is dealt


with, and otherwise c3 is equal to zero. Similarly
for c2 if a 2-phase ac machine is used, likewise for
cl in the case of 1-phase or DM models. It is to be
pointed out that the inverse of FT(t)1 is to be obtained by transposing the rP(0)1 and ESI matrices.

rT(t) 1

1-phase and DM models.

TAGS

Electric

Fig. 3.

Inf ini te

controll

mass

dynamic

Excitation
-UMcapacitors

Power
system

Electricall, equivoalent

Network

(EN)

Illustrative wind energy conversion system.

1347
significant in quantity, note that none of this TACS
modeling directly affects the UM (each coil of which
could be directly excited by a TACS variable). Thus
our example does not demonstrate at all the power and
flexibility of the TACS/UM interface, unfortunately.
Summary statistics of the TACS modeling are as follows:
3 function blocks, 22 summers, 26 TACS sources, and 60
supplemental variables.

EMTP simulations of this problem using BPA's DEC


computer system required approximately 2200
seconds of computer time for each second of model time,
during a severe transient. A time step of .050 msec
was used for all simulation without difficulty, meaning
that 20 000 steps were taken each model second. EMTP
electrical network data were characterized by summary
statistics which follow:
86 nodes, 113 branches, 31
switches, and 383 nonzero upper-triangular entries of
the factored admittance matrix.

VAX-11/780

Fig. 4.

Electrical analog representing mechanical


system.

The branch currents and


"generator", respectively.
voltages of this equivalent represent shaft torques and
angular speeds, respectively.
Capacitance ("C") to
ground represents a mechanical moment of inertia,
inductance ("L") corresponds to the reciprocal of a
mechanical spring coefficient, and resistance ("R")
represents the reciprocal of a mechanical damping
coefficient.
Blade speeds are represented by timedependent current sources which dynamically relate the
torques, speeds, and angles of the blades to the ambient wind speed E121. There is coupling between this
mechanical modeling and the electromagnetic portion of
the generator via current source TUM (electromagnetic
torque) and node voltage WUM(speed of the rotor).

Recent studies r12] using the MOD-2 WTS with a


synchronous generator have demonstrated the important
influence of the "soft" low-speed shaft on the critical
clearing time (following a fault). By definition, the
fault must be cleared within this time, in order to
avoid a loss of synchronism. The soft shaft is added
to keep torques within design limits, though it seems
to result in undesirably-short critical clearing times.
The illustration of this section uses an induction
rather than a synchronous generator, in an attempt to
drastically improve this crucial statistic.

First consider energization of the WTS, starting


from rest at time zero. The build up of mechanical
speed, electromagnetic torque, and terminal voltage of
the induction generator is shown in Fig. 5. There is
no connection to the main power grid at the start, so
it is residual magnetism of the induction machine which
provides the initial excitation. The rectifier and the
inverter are only activated once the generator has
developed rated voltage, at times'2.5 and 2.6 seconds.
Small fluctuations in the torque at this point can be
clearly observed, and they are important for the study
of machine losses. Correct convergence to the steadystate operating condition is shown by Fig. 6, which
documents the behavior of inverter variables for a
firing angle of 135 degrees. In order, one sees the
a-phase forward thyristor voltage and current, forward
thyrister currents for all three phases, the generator
a-phase terminal voltage, and finally, the current of
the dc link. It might be remarked that the abrupt
-changes in the thyristor currents are due to choice of
the time step (.050 msec). If one wants to observe the
rapid oscillations of the snubber circuits, this should
be decreased. That all other problem variables are

Two UM modules are used in the representation of


That on the left, connected to MOD-2, is set
Fig. 3.
to be an induction generator with a cage rotor.
Crucial to this application is the UM representation of
residual flux (for starting) and magnetic saturation
(for stable steady-state operation). The second UM
module (on the right) represents a 3-phase synchronous
machine, as an approximation to a strong connected
power grid. The moment of inertia of this SM is chosen
very large, so that frequency deviation will be small.
Electric network modeling in Fig. 3 consists of
important excitation capacitors, the two thyristor
bridge circuits (3-phase, full-wave power electronic
converters), associated snubber networks and smoothing
reactors, the dc transmission link, the passive R-L
load, and the ac transmission line which connects with
the power network. It should be mentioned that part of
the transmission line is represented by distributed
modeling E1i, in order to decouple the two UM usages of
compensation (a restriction of the present coding).

1.1

Finally, the TACS modeling of Fig. 3 is used for


of the thyristor firing angles, as well as for
modeling of the blade dynamics (which require special
analytical relationships, and also the insertion of the
While
nonlinear power coefficient distribution r121).
control

2.2

b) Electromagnetic torque of IM,

the

Time (sec)

TUM (*1.E4)

h
4

0-

c) Envelope of terminal voltage, vab (*10 kV)


Fig. 5.

Energization of wind generator.

1348
2-

:n.

a) Forward valve voltage, v+

same

valve,

TUM (*1.E4)

4.

a.2

b) Current in

a) Electromagnetic torque of IM,

(*10 kV)

i+2 (*100 A)

5.333
b) Current in DC link,

5.05

iDC

5.617

(*100 A)

Time (sec)

6.300

6.2755
c) Torque of soft shaft connecting
blade 1 with hub, TBLH (*1.E6 N-m)
4.7500

4.6875

d) Terminal voltage of IM,

vab

d) Torque in soft shaft connecting the


hub and the generator, THG (*1.E6 N-m)

(*10 kV)

4'-

Ad

e) Current in
Fig.

6.

dc

link, iDC (*100 A)

Steady-,state operation

at

rated

valid in spite of our large time step is


to the stability of the trapezoidal rule

load.

testamonial

rni.

Next, consider the worst electrical disturbance


a solid 3-phase fault at the terminals of
imaginable:
the induction machine. Fig. 7 displays the transients
associated with a fault initiation at time 5.100
seconds, and fault clearing 20 cycles later.
It is
seen that recovery of the voltage and torque require an
additional 100 msec or so, as the generator rebuilds
the needed excitation flux. There is indeed complete
recovery from this long and severe fault, however,
without any need for special resynchronization. This
demonstrates a most important advantage of induction
generators (compared with their synchronous relatives).
Fig. 7 also shows the expected salubrious influence of
the soft shaft on torques TBlH and THG

~~~~AAAAAhHih

e) Positive half of terminal voltage


envelope of IM, vab (*10 kV)
Fig. 7.

Three-phase fault at 5.1 sec, with


clearing after 20 cycles.

Additional studies of the sample system are both


interesting and necessary, for a complete engineering
analysis. However, they are beyond the scope of our
present limited exposition.

The numerical efficiency of UM modeling could be


considered, as a final aspect. Users might worry that
the extreme flexibility of this modeling be paid for by
seriously-degraded numerical efficiency. However, no
substantiation of such fears has yet- been observed
(though comparisons are far from complete). The SM in
Fig. 3 is quite conventional, and it really does not
require the new UM for representation. Alternatively,
special 3-phase synchronous machine models have existed
for several years, as originally developed for studies
of SSR r2, 131. Replacement of the SM in Fig. 3 by

1349

either the Type-50 r21 or Type-59 r141 EMTP synchronous


machine models F151 was found to have little effect on
total problem running time, for two reasons.
First,
there is enough other modeling (e.g., TACS and valves)
so that numerical effort associated with the machinery
does not dominate. Second, UM efficiency does seem to
compare favorably (for this test, UM running times were
bounded by those for the other two alternatives).
Thusfar, numerical efficiency of the UM has been a most
pleasant surprise (it was not Xa design requirement).

-nmP

IMi

-R P~ -UMP - d(X-U!NP )/dt


:

UM
um

w=

and if

YUMP
YUM

+1

0o

-1

(
.col
>d

TUM

UM

X, then

(Y

=col

(yOE Yd I Yd2

,Y

,y

FR1
E

diag (R

diag (R OE

,R
dl
R

ql

Ydh

fAp, (t)1

R..
q2

X UMP

,rLP 1

are

UMP

dN

N
+ qz

Nq

+ (L 2d4

..

d h

Ii
idiq

(A4)

(t-At)

ht-E

-UME ( )

(t)1X (t)

rA

rFA_ lv

A; ()1-m

(A5)

(t)

(t)

(A6)

Ev UME

with:

rUt3 - w(t)rFpirGp 1NlrL 1


= FFPFrGpI fw (t)FNl - 2FU31/At I

rAp, (t)1
FApX(t)

rA ( )1

rFAE

GFE1FCE 1

-1

FGP1 = FRP
FFp1

2FFEIGEj/At

iApvi -1rFpFlGp1

Fa 1 = FR 1

IrFL,P 1/At

= ( Fu31 + 2rF

rFE1 = ( IU1

2rGF FrL I/At )

hE(t-At)

is

to

UMP(t-At) +~+ FV1x

(t-At)
At)
A.v Z(t-

ME

(t-At)

+
+

rAEV-1X
1v(t-At)
Ev-E

mE(t-At)

where

rFp IfGq 112FU3 I/At + w(t-At)rN1}FkP


- ~rFE + 2FFE 1rEG/At
EF FrG 112FU 1/At - w(t-At)FN1}

1 = -

qk

related to the flux linkages by


X mP

m
iq

iX
dmXq

(t)

h (t-At) = r1

Note that the total number of exciataion coils is eqixal


to 1+h+k

The currents

I qmd
qX

= n
p

hE(t-At) = FrH 1i
.....

Lkq k

( id

qk

*,Rd2

mqq

The history functions h p(t-At) and


be found from

dh

rGEI

rotor angle

yq2

.....

Lmd

mq

.
.....

It is to be noted that the needed inversion to find the


rEEl, and FFl1 need not to be done
matrices
F
1,.?1,
by the computer. Their inverse is immediately found by
reciprocating each entry due to the fact that the conductance matrices
and FGp1 are diagonal. To solve
the UM output, i.e. all UM currents, the only inversion needed is for the (3x3)-matrix FA 1.

speed

=col

yq

Rp

inmd

mnd

d2

LPq 2.

Equations (Al) and (A2) are differential equations


which have to be converted to recursive equations for
the purpose of digital computations. The numerical intepration routine chosen here is the Trapezoidal Rule,
which applied to (Al) and (A2) yields:

(A2)

angular

rnq

Finally the following expression of the electromagnetic torque as produced by the UM completes the
electromagnetic system equations of the UM:

(Al)

,= n

or

X
wFNl.,-UMP

>. =I

d(e)/dt

v or

d(X UE) /dt

Ei

FNl

nmd

Liql

The UM voltage equations of respectively the power


side and excitation side for a reference system stationary with respect to the excitation side can be formulated as follows

wi th

1 = diag (Lo, LQ , L )
kp op
kd
91c~~~~~,q
= col (
, A , X )

FL QE l diag
L d I ,L
d (L
OE
L(O

APPENDIX 1

FR

(A3b)

--WE

mq

The demand for transient network simulation which


includes a wide variety of doubly-fed rotating electric
machines has now been satisfied by the development of
universal machine (UM) logic. Included in this digital
computer modeling are all major classes of electric
generators and motors that are in common use today.
Via a simple selection-of binary control parameters,
any one machine of interest can be selected from the
universal model. Further, UN logic has been integrated
with BPA's EMTP, to provide the required additional
modeling for control systems, the electric network, and
connected mechanical systems (by electrical analog).
Particular care has been taken to maximize flexibility
of the UM interface, with
compensation used for
the connection of both the mechanical shaft and all
electrical coils. As a consequence, it is now possible
to represent various sophisticated extensions to the
classic machine models (e.g., frequency-dependent rotor
resistance for induction machines). Finally, usage has
been illustrated by the simulation of a proposed wind
energy system (based on a capacitor-excited induction
generator and the NASA MOD-2 wind turbine).

-V-UME

CONCLUSIONS

7.

vUNP
-iUN

FL QE I -UME
i

-UME
with: FL

(A3a)

nh 1

FH 1

rH

Ex

rp 1

1 = fF 1FG 1
E

1350

REFERENCES

r,i H.

W. Dommel, "Digital Computer Solution of Electromagnetic Transients in Single and Multiphase


Networks." IEEE Transactions on Power Apparatus and
Systems, vol. PAS-88, pp. 388-399, April 1969.
[2] G. Gross and M. C. Hall, "Synchronous Machine and
Torsional Dynamics Simulation in the Computation of
Electromagnetic Transients." IEEE Transactions on
Power Apparatus and Systems, vol. PAS-97, pp.
1074-1086, July/August 1978.
E31 D. C. White and H. H. Woodson, "Electromechanical
Energy Conversion." New York: John Wiley, 1959,
pp. 254-345.
141 R. H. Park, "Two-Reaction Theory of Synchronous
Machines, Generalized Method of Analysis." AIEE
Transactions, vol. 52, pp. 352-355, 1933.
[5] L. Dube and H. W. Dommel, "Simulation of Control
Systems in an Electromagnetic Transients Program
with TACS." IEEE PES Power Industry Computer Applications (PICA) Conference Record, vol. 10, pp.266271, 1977.
16 J. Meisel, "Principles of Electromechanical Energy
Conversion." New York: McGraw-Hill, 1966, pp. 1-43.
[77 W. F. Tinney, "Compensation Methods for Network
solutions by Triangular Factorization." IEEE Transactions on Power Apparatus and Systems, vol.
PAS-91, pp. 123-127, Jan/Feb 1972.
181 P. M. Anderson and A. A. Fouad,
"Power System
Control and Stability." Ames, Iowa: Iowa State
University Press, vol. 1, pp. 83-85, 1977.
19] J. R. Smith, G. J. Rogers, G. W. Buckley, "Application of Induction Motor Simulation Models to Power
Station Auxiliary Pump Drivers." IEEE Transactions
on
Power Apparatus and Systems, vol. PAS-98,
pp. 1824-1831, Sept/Oct 1979.
[iol N. Mohan and M. Riaz, "Wind-Driven, CapacitorExcited Induction Generators for Residential Electric Heating." Paper No. A78 050-7 presented at
the 1978 IEEE PES Winter Meeting in New York City.
1111 N. G. Hingorani and J. Slapp, "Application of Induction Generators and DC Transmission Links to
Small Hydroelectric Power Plants." Madison, Wisconsin: University of Wisconsin-Extension, Dept.
of Engineering, Proceedings of symposium on HVDC
Power Transmission as given in Phoenix, Arizona,
March 24-27, 1980.
1121 0. Wasynezuk, D. T. Man, I. P. Sullivan, "Transient Behaviour of Wind Turbine Generators During
Electrical Disturbances." Paper No. 81 WM 173-4
presented at the 1981 IEEE PES Winter Meeting in
Atlanta, Georgia.

r13l A.

G.

Phadke, Organizer,

Electrical

Course

Transient

"Digital Simulation of

Phenomena."

No. 81 EH0173-5-PWR.

IEEE

Tutorial

r141 V. Brandwajn

and H. W. Dommel, "Simulation


of
Generators in Electromagnetic Transients
Programs." International Journal of Electrical

Turbine

Energy Systems, vol. 1, no. 2, pp.


118-124, July 1979.
W. S. Meyer and T. H. Liu, Eds., "EMTP Rule Book."
Portland, Oregon: Bonneville Power Administration,
System Engineering, 1980, 746 pages.
Power and

r151

W.
Scott Meyer (M'69) was
born in Madison, Minnesota in 1942.
He received the B.S., M.S.E.E., and
Ph.D.
degrees from the University
of Minnesota in 1964, 1966, and

1969, respectively.

From 1969 to 1971, Dr. Meyer


taght power system engineering at
the University of Minnesota in Min-

neapolis as an Assistant Professor.


He
then

spent almost

year

with

Systems Control, Inc. (Palo Alto, California) before


moving to his present position with the Bonneville
Power Administration in Portland, Oregon.
Since mid-1973, the computer solution of electromagnetic transient problems has been an all-engrossing
preoccupation.
Hian K.
Lauw was born in
Indonesia in 1943. He received the B.S.
M.S.E.E.
and
Ph.D.
degrees from the Delft University of Technology, the Netherlands, in 1966, 1968 and 1977 respectively. From 1968 till 1978 he
was a scientific associate to the
Laboratory of Electric Machinery
and Transformers in Delft. He then
joined the Department of Electrical
and Computer Engineering, Oregon
State University, Corvallis,
in

Bogor,

September 1978.

His field of interest is dynamics of electric machines, power electronics and stability analysis of
large-scale power systems.

Discussion
Jaime Avila-Rosales (National Polytechnic Institute of Mexico, and
Electric Power Research Institute of Mexico): The authors are to be
congratulated for developing a Universal Model for rotating electric
machinery in the BPA'S EMTP.
I would appreciate the authors clarification on the following:
a) Could the authors give results of a simulation in which the eddy
current losses in the excitation portion of the machine are shown
to be important?
b) Have the authors included in their model the effect that frequency
can have on the eddy current losses? If so, could they give an explanation of the way they have done it?
The last question if concerned with the effect that eddy currents in solid
iron and dampers may have when frequency varies.
Manuscript received August 18, 1981.
H. K. Lauw and W. S. Meyer: We wish to thank Dr. Avila-Rosales for
his interest in the paper. Taking into account of eddy current losses is
important to transient studies involving unbalanced loads, faults or
energization. The physical phenomenon is extremely complicated due
to the frequency effect. A common strategy is to determine by frequency response techniques the parameters of additional short-circuited coils
which are employed to approximate the result of the phenomenon.
Once that these parameters are established, then the coils can simply be
included in the basic machine model due to the variable dimensioned
coil structure of the UM. For a clear discussion on the application of
this approximation method to induction machines, we refer to
Reference 9 of the paper. For synchronous machines, Reference 16 is

appropriate.

1351
By using the EMTP, another approach is possible if the frequency
dependence of the relevant machine parameters (resistance, inductance)
due to the eddy currents is known. The frequency dependent parameter
is inserted in the Electric Network Module of the EMTP; interface with
the UM module is taken care of automatically. As a result the solution
feature with frequency dependent parameter as developed for the Electric Network Module is readily available to the user. Conceptionally the
method is based on applying convolution theory to obtain solutions in
the time domain from the frequency domain. Numerical realization involves the collection of multiple past history points of the solution
which are weighted with a convolution integral. For a detailed account
we refer to References 17, 18 and 19.
REFERENCES

[16] G. J. Rogers and J. R. Smith, "Synchronous Machine Model Including Eddy Currents", Proc. IEEE, Vol. 120, pp. 468, 1973
[17] W. S. Meyer and H. W. Dommel, "Numerical Modeling of
Frequency-Dependent Transmission Line Parameters in an Electromagnetic Transients Program", IEEE Trans. on Power App.
Syst., Vol. PAS-93, pp. 1401-1409, 1974

[18] A. Semlyen and A. Dabuleanu, "Fast and Accurate Switching


Transient Calculations on Transmission Lines with Ground
Return Using Recursive Convolutions", IEEE Trans. on Power
App. Syst., Vol. PAS-94, pp. 561-571, 1975
[19] A. Ametani, "A Highly Efficient Method for Calculating Transmission Line Transients", IEEE Trans. on Power App. Syst., Vol.
PAS-95, pp. 1545-1551, 1976
Manuscript received September 22, 1981.

Você também pode gostar