You are on page 1of 6

On the state space representation

of synchronous generators
Emmanuel Delaleau

Université Paris-sud
Laboratoire des signaux et systèmes
CNRS – Supélec
3 rue Joliot-Curie
91 192 Gif-sur-Yvette, France
e-mail: Emmanuel.Delaleau@lss.supelec.fr

Abstract
The paper aims to present a formalism allowing the selection of state variables
in dymamic models. This is applied to the model of a synchronous generator result-
ing in a new form of the state equations not encountered in the existing litterature.
The paper will present the properties of this new model in terms of simulation, sta-
bility and control, and compares it to more conventional state-space models. The
main difference relies in the fact that the model presented here does not include
any load which are externeal to the generator. This should have consequences for
simulation softwares.

1 State-variable representation of nonlinear systems


In this paper we present a recent formalism for selecting state variables in dynamic
models of electromechanical systems. We illustrate the procedure on the example of
a synchronous generator. The procedure is useful in control and estimation of elec-
tromechanical systems, and if offers an alternative to the conventional intuition-based
modeling process.
A state-variable representation of a nonlinear system generally takes the form:

ẋ = f (x, u, $) (1a)
y = h(x, u, $) (1b)

where u is the control vector, $ the disturbance vector, x the state vector and y is the
output vector. In some cases, f and h can explicitly depend on time and, usually, the
output equation (1b) is only a function of the state and does not depend on u nor on $.
The input in (1), which consist of u and $, can be thought as the “cause” acting
on the system. Mathematically the vector functions t 7→ u(t) and t 7→ $(t) must be

1
specified or known in order to calculate or simulate a trajectory of the underdetermined
system of equations (1).
State variables (i.e. components of any state vector) are abstract mathematical
quantities, useful to obtain the first-order representation (1) when the input is known.
However, in most of the cases state variables are chosen among the physical variables
representing the plant under study.
Any output y can be thought as the “effect” of the operation of the system. It is
usually formed with the “to-be-controlled” variables. Most of the time, its components
correspond to measured variables.
A recent framework of nonlinear control [2] resulted in a more general state-variable
representation in which input derivatives may appear:

ẋ = f (x, u, $, u̇, $̇, . . .) (2a)


y = h(x, u, $, u̇, $̇, . . .) (2b)

In addition to the mathematical foundation of this type of equations, some examples


of engineering systems have been found to belong to this class of say “generalized
systems” (see [4] for the case of an overhead crane).
A careful discussion of choice of input and output in terms of cause and effect
about electrical machines leads to the conclusion that the motor operation can be rep-
resented by (1) whilst the generator operation only admits state-variable representation
of type (2). More precisely, natural outputs of the electrical generators let appear input
derivatives. The paper will examine the obtained generalized state model of a syn-
chronous generator in perspective of stability, simulation, and control.

2 Introductory examples
The following two examples —namely the DC and synchronous permanent magnets
machines— are use to introduce our point as they are quite simple and easy to under-
stand. Thought this two machines are rarely used as generators, it is interesting to look
at their state space models in this mode of operation. More practical and complicated
cases can be analysis with the same tools.

2.1 The permanent magnets DC machine


The basic differential-algebraic equations that models the behavior of a permanent
magnet DC machine are:

J Ω̇ = Kt I − Kf Ω − Tex (3a)
dI
U = L + RI + Kb Ω (3b)
dt
where  = +1 in motor operation and  = −1 in generator operation. The variables
are Ω (angular speed), I (current), U (voltage) and Tex (external torque applied to
the shaft). The parameter are J (momentum of inertia), Kt (torque constant), Kf

2
(friction coefficient), L (coefficient of auto-inductance), R (resistance) and K b (back
emf constant).
Notice this is a linear system and this is a 2-inputs system. The main reason is that
this is a system of 2 independent equations relating 4 variables —namely Ω, T ex , I
and U . So one needs to provide 2 time functions to be able to express its solution or
to simulate it. (see [3] for the details about the notion and size of the input of a linear
system.)

Motor operation. In this situation, the natural control input is u = (U ) and the
disturbance input is $ = (Tex ). Note that in this case, Tex represents the load torque
undergone by the schaft. Knowing the time functions t 7→ U (t) and t 7→ T ex (t)
(specification of the input), one needs 2 initial conditions to express the solutions of (3).
(See also [3] for an intrinsic determination of the dimension of the state.) A natural
choice of state is thus x = (Ω I)t . The output is e.g. y = (Ω), leading to the classical
or Kalman state representation:
     
−Kf /J Kt /J 0 −1/J
ẋ = x+ u+ $
−Kb /L R/L 1/L 0
y = (1 0) x

Generator operation. In this case, the control input is the torque applied to the motor
u = (Tex ), which here represents the torque applied to drive the generator, and the
disturbance input is the electrical load of the generator represented by the current $ =
(I). (Remember that in this case  = −1.) For a given behavior of t 7→ T ex (t) and
t 7→ I(t), one needs only one initial condition to express the solution of (3). There
is only one state variable to pick and a natural choice is x = (Ω). The output is in
this mode of operation clearly the voltage produced by the generator y = (U ). The
corresponding representation reads:
Kf 1 Kt
ẋ = − x− u+ $
J J J
y = Kb x + R$ + L$̇

which let appear the disturbance (input) first derivative in the output equation.

2.2 The permanent magnets synchronous machine


For the sake of simplicity consider a non salient pole machine such that L = L d = Lq
whose model in the DQ frame reads:

θ̇ = Ω (4a)
J Ω̇ = np Km iq − Kf Ω − Tex , (4b)
did
L = −Rid + np LΩiq + vd (4c)
dt
diq
L = −Riq − np LΩid − np Km Ω + vq (4d)
dt

3
where  = ±1 whether one considers motor or generator operation, i d and iq are
called respectively the “direct” and “quadrature” currents, v d and vq are the direct and
quadrature voltages.
As this model contains 4 equations relating 7 variables, this in a 3 input system.

Motor operation. In this case, the input is u = (vd vq )t (control) and $ = Tex
(disturbance). This choice being done it is not difficult that ones need 4 initial condition
to express the behavior of (4) in motor mode. The natural choice of state variable is
x = (θ Ω id iq )t . The output can be for instance y = (θ) or y = (Ω) according to the
type of control task is to be achieved.
So, the state representation is ( = +1):

ẋ1 = θ̇ = Ω (5a)
np Km Kf 1
ẋ2 = Ω̇ = iq − ω− Tex (5b)
J J J
did R 1
ẋ3 = = − id + np ωiq + vd (5c)
dt L L
diq R np Km 1
ẋ4 = = − iq − np ωid − ω + vq (5d)
dt L L L
y = θ (5e)

which is of type (1).

Generator operation. In this mode of operation, the control input is the torque ap-
plied to the machine u = (Tex ) and the disturbance is represented by the currents
$ = (id iq )t . Consequently one need only 2 initial condition to express the behavior
of the generator mode, and a natural choice is x = (θ Ω). Obviously, the output is
y = (vd vq ).
The state variable representation is then ( = −1):

ẋ1 = θ̇ = ω (6a)
Kf 1 np Km
ẋ2 = ω̇ = − ω + Tex − iq (6b)
J J J
did
y1 = vd = Rid − np Lωiq + L (6c)
dt
diq
y2 = vq = Riq + np Lωid − np Km ω + L (6d)
dt
Note that the (disturbance) input derivatives appears in the output equations.

3 State variable model of the synchronous generator


The final paper will expose the development of the state-space model of the syn-
chronous generator (with field excitations and damping windings) and its consequence
for simulation and control.

4
A model of such a electrical generator can be expressed in an appropriate frame by:

did diF diD


vd = rid + ωLq iq + ωkMQ iQ − Ld − kMF − kMD (7a)
dt dt dt
diq diQ
vq = −ωLd id − ωkMF iF − ωkMD iD + ri q − Lq − kMQ (7b)
dt dt
did diF diD
vF = −rF iF + kMF + LF + MR (7c)
dt dt dt
did diF diD
0 = rD iD − kMD − MR − LD (7d)
dt dt dt
diQ diQ
0 = rQ iQ − kMQ − LQ (7e)
dt dt
J ω̇ = Cm − Cem − Dω (7f)
Cem = 1/3 [(Lq − Lq )id iq + kMF iq iF + kMD iq iD − kMQ id iQ ] (7g)

For the details we refer the reader to [1].


The natural choice of control input is u = (Cm vF )t and the disturbance is in this
case $ = (id iq )t . The behavior of the generator is modeled by a 4th order dynamics
and a natural choice of state is thus x = (ω iF iD iQ )t . The output is obviously the
to-be-regulated windings voltages y = (vd vq )t .
The final paper will discuss precisely the derivation of the state space model which,
as the simpler previous examples, also involve disturbance derivatives in the output
equations. Moreover, one will provide a precise discussion of the consequences in
terms of simulation, control and stability of the generator. In particular one will proceed
in a careful comparison with more conventional models, in which the energy-storage
variables are the states.
The main difference of the approach adopted here with more conventional works is
that we only consider the generator itself in the modeling process and not the electrical
load, which can be of various type. The load must be of course taken into consideration
in simulation of the behavior of the generator but it is preferable to include a model of
the load in a separated “box”. This is important in “stand-alone” applications but also
in most of the simulation software that consider every part of a complex ssytem as an
“elementary box”. To our point of view, the model of any generatro does not need to
include any part of the load. The result of the paper should have consequence in the
development of simulation softwares.

References
[1] P. M. Anderson and A. A. Fouad. Power Systems Control and Stability. Iowa State
University Press, 1977.

[2] M. Fliess. Generalized controller canonical forms for linear and nonlinear dynam-
ics. IEEE Trans. Automat. Control, 35:994–1001, 1990.
[3] M. Fliess. Some basic structural properties of linear generalized systems. Systems
Control Lett., 15:391–396, 1990.

5
[4] M. Fliess, J. L´evine, and P. Rouchon. A generalized state variable representation
for a simplified crane description. Internat. J. Control, 58:277–283, 1991.

You might also like