Sei sulla pagina 1di 14

198 Power systems electromagnetic transients simulation

u y +

e
G
1
G
s
Figure 8.3 First-order lag
written in FORTRAN, C, C++, MATLAB or BASIC, and link this with EMTDC.
Interpolation of controls is also an important issue for accurate results [6].
8.3.1 Example
The rst-order lag control system, depicted in Figure 8.3, is used to demonstrate the
use of the z-domain for the prediction of instabilities.
The corresponding transfer function is:
y
u
=
1
1 +fg
=
G
1 +s
(8.1)
where
f = feedback path = 1/G
g = forward path = G/(s)
= time lag.
The equations for the two blocks are:
e = u
1
G
y (8.2)
y =
Ge
s
(8.3)
Substitution of the trapezoidal rule to form difference equations gives:
e
k
= u
k

1
G
y
k
(8.4)
y
k
=
t (e
k
+e
k1
)G
2
(8.5)
The difference in data paths becomes apparent. If solved as two separate difference
equations, then e
k
must be calculated from y at the previous time step as y
k
is not
available and this introduces one time-step delay in the y data path. Swapping the
order of equations will result in the same problemfor the e data path. Substituting one
equation into the other and rearranging, results in a difference equation with no delay
Control and protection 199
in data path. This is equivalent to performing integrator substitution on the transfer
function for the complete controller
Time-step delay in data path
If there is a time-step delay in the feedback path due to the way the difference equation
for each block is simulated, then
e
k
=
_
u
k

1
G
y
k1
_
y =
G
s
e
(8.6)
Applying trapezoidal integration gives:
y
k
= y
k1
+
t G
2
(e
k
+e
k1
)
= y
k1
+
t G
2
_
u
k

1
G
y
k1
+u
k1

1
G
y
k2
_
=
_
y
k1

t
2
y
k1

t
2
y
k2
_
+
t G
2
(u
k
+u
k1
) (8.7)
Transforming equation 8.7 into the z-plane yields:
Y
_
1 z
1
_
1
t
2
_
+z
2
t
2
_
=
t G
2
(1 +z
1
)U (8.8)
Rearranging gives:
Y
U
=
(t G/(2))(1 +z
1
)
_
1 z
1
(1 t /(2)) +z
2
t /(2)
_
=
(t G/2)z(z +1)
_
z
2
z
1
(1 t /(2)) +t /(2)
_ (8.9)
The roots are given by:
z
1
, z
2
=
b

b
2
4ac
2a
(8.10)
z
1
, z
2
=
1
2

_
1
t
2
_

_
_
1
t
2
_
2
4
t
2

=
1
2

_
1
t
2
_

_
_
1 3
t

_
+
t
2
4
2

(8.11)
Stability is assured so long as the roots are within the unit circle |z| 1.
200 Power systems electromagnetic transients simulation
No time-step delay in data path
If there is no delay in the feedback path implementation then e
k
= (u
k
(1/G)y
k
)
Applying trapezoidal integration gives:
y
k
= y
k1
+
t G
2
(e
k
+e
k1
)
= y
k1
+
t G
2
_
u
k

1
G
y
k
+u
k1

1
G
y
k1
_
=
_
y
k1

t
2
y
k

t
2
y
k1
_
+
t G
2
(u
k
+u
k1
) (8.12)
Transforming equation 8.12 into the z-plane yields:
Y
__
1 +
t
2
_
+z
1
_
t
2
1
__
=
t G
2
(1 +z
1
)U (8.13)
Rearranging gives:
Y
U
=
(t G/(2))(1 +z
1
)
(1 +t /(2)) +z
1
(t /(2) 1)
=
(t G/(2))(z +1)
z(1 +t /(2)) +(t /(2) 1)
(8.14)
The pole (root of characteristic equation) is:
z =
(1 t /(2))
(1 +t /(2))
(8.15)
Note that |z
pole
| 1 for all t /2 > 0, therefore this method is always stable.
However this does not mean that numerical oscillations will not occur due to errors
in the trapezoidal integration.
Root-matching technique
Applying the root-matching technique to this control system (represented mathemat-
ically by equation 8.1) gives the difference equation:
Y(z)
U(z)
=
G(1 e
t /
)
(1 z
1
e
t /
)
(8.16)
hence multiplying both side of equation 8.16 by U(z)(1 z
1
e
t /
)
Y(z) = e
t /
z
1
Y(z) +G(1 e
t /
)U(z) (8.17)
Transforming to the time domain yields the difference equation:
y
k
= e
t /
y
k1
+G
_
1 e
t /
_
u
k
(8.18)
Control and protection 201
The pole in the z-plane is:
z
pole
= e
t /
(8.19)
Note that |z
pole
| 1 for all e
t /
1 hence for all t / 0.
Numerical illustration
The rst-order lagsystemof Figure 8.3is analysedusingthe three difference equations
developed previously, i.e. the trapezoidal rule with no feedback (data path) delay, the
trapezoidal rule with data path delay and the exponential formusing the root-matching
technique. The step response is considered using three different time-steps, t =
/10, , 10 ( = 50 s) and the corresponding results are shown in Figures 8.48.6.
When t / = 1/10 the poles for trapezoidal integration with delay in the data
path are obtained by solving equation 8.11; these are:
z
1
, z
2
=
19
20

_
_
_
19
20
_
2

4
20
_
= 0.0559 and 0.8941
Since two real roots exist (z
1
= 0.0559 and z
2
= 0.894) and both are smaller than
one, the resulting difference equation is stable. This can clearly be seen in Figure 8.4,
which also shows that the exponential form (pole = 0.9048) and the trapezoidal
rule (pole = 0.9048) with no data path delay are indistinguishable, while the error
introduced by the trapezoidal rule with data path delay is noticeable.
0
0.2
0.4
0.6
0.8
1
1.2
y
Time (s)
Trapezoidal Trapezoidal (time-step delay) Exponential form
0.0009 0.001 0.0011 0.0012 0.0013
Figure 8.4 Simulation results for a time step of 5 s
202 Power systems electromagnetic transients simulation
y
Time (s)
Trapezoidal Trapezoidal (time-step delay) Exponential form
0 0.001 0.002 0.003 0.004 0.005
0
0.35
0.7
1.05
1.4
Figure 8.5 Simulation results for a time step of 50 s
y
Time (s)
Trapezoidal Trapezoidal (time-step delay) Exponential form
0
400
800
1200
400
0 0.001 0.002 0.003 0.004 0.005
Figure 8.6 Simulation results for a time step of 500 s
Control and protection 203
The poles when t / = 1 are given by:
z
1
, z
2
=
1
2

1
2

_
_
_
1
2
_
2
4
1
2
_

= 0.25 j0.6614
Hence a pair of complex conjugate roots result (z
1
, z
2
= 0.25 j0.6614). They
lie inside the unit circle (|z
1
| = |z
2
| = 0.7071 < 1) which indicates stability.
Figure 8.5 shows that, although considerable overshoot has been introduced by the
time-step delay in the data path, this error dies down in approximately 20 time steps
and the difference equations are stable. A slight difference can be seen between the
trapezoidal integrator with no data delay (pole = 0.3333) and the exponential form
(pole = 0.3679).
Finally when t / = 10 the poles for the trapezoidal rule with time delay in the
feedback path are:
z
1
, z
2
=
4
2

_
((4)
2
4 5)
2
= 2.0 j1.0
hence two complex poles exist, however they lie outside the unit circle in the z-plane
and therefore the system of difference equations is unstable. This is shown in the
simulation results in Figure 8.6.
The poles for the trapezoidal method with no time delay and exponential form are
0.6667 and 4.5400e005 respectively. As predicted by equation 8.15, the difference
equation with no data path delay is always stable but close examination of an expanded
view (displayed in Figure 8.6) shows a numerical oscillation in this case. Moreover,
this numerical oscillation will increase with the step length. Figure 8.6 also shows
the theoretical curve and exponential form of the difference equation. The latter gives
the exact answer at every point it is evaluated. The exponential form has been derived
for the overall transfer function (i.e. without time delays in the data paths).
If a modular building block approach is adopted, the exponential form of dif-
ference equation can be applied to the various blocks, and the system of difference
equations is solved in the same way as for the trapezoidal integrator. However, errors
due to data path delays will occur. This detrimental effect results fromusing a modular
approach to controller representation.
This example has illustrated the use of the z-domain in analysing the difference
equations and data path delays and shown that with z-domain analysis instabilities
can be accurately predicted. Modelling the complete controller transfer function is
preferable to a modular building block approach, as it avoids the data path delays
and inherent error associated with it, which can lead to instabilities. However the
error introduced by the trapezoidal integrator still exists and the best solution is to
use instead the exponential form of difference equation derived from root-matching
techniques.
As well as control blocks, switches and latches, the PSCAD/EMTDC CSMF
library contains an on-line Fourier component that is used to derive the frequency
components of signals. It uses a Discrete Fourier Transform rather than an FFT. This
204 Power systems electromagnetic transients simulation
1.0
Magnitude
Phase
Freq
Mag
sin
0.0
Phase
50.0
Frequency
0
.
0
0
1
1
2
D
0
.
0
0
1
D
4
2
D
2
2
D
1
0
0
0
.
0

0
.
1
V_load
I_load
I_load
V_load
GTO_AP
GTO_AN
12
0
.
0
12
0
.
0
Vs1
Vt
Monitoring of a load
GTO_AP
GTO_AP
GTO_AN
GTO_AN
3
2
A
B
Compar-
ator
+

Figure 8.7 Simple bipolar PWM inverter


1.0
Magnitude
Phase
Freq
Mag
sin
0.0
Phase
50.0
Frequency
Vs1
Vt
Vt
Vs1
GTO_APx
1
2
Vt
Vs1
When the signal H(OFF) becomes larger
than L(OFF) it creates the OFF signal (0)
L
H
2
H
L
ON
OFF
When the signal H(ON) becomes larger
than L(ON) then creates the ON signal (1)
Amount (seconds) between time
points switching occured.
Main output (1: ON, 0: OFF)
Figure 8.8 Simple bipolar PWM inverter with interpolated turn ON and OFF
Control and protection 205
enables a recursive formulation to be used, which is very efcient computationally,
especially when a small number of frequency components are required (the maximum
limit is 31 for the on-line Fourier component).
Figure 8.7 displays a simple bipolar inverter, where the valve is turned on at
the rst time point after the crossover of the triangular and control signals. To take
advantage of interpolated switching the interpolated ring pulses block is used as
shown in Figure 8.8.
8.4 Modelling of protective systems
A protective system consists of three main components, i.e. transducers, relays and
circuit breakers, all of which require adequate representation in electromagnetic
transients programs.
The modelling of the relays is at present the least advanced, due to insufcient
design information from manufacturers. This is particularly a problem in the case
of modern microprocessor-based relays, because a full description of the software
involved would give the design secrets away! Manufacturers manuals contain mostly
relay behaviour in the formof operating characteristics in terms of phasor parameters.
Such information, however, is not sufcient to model the relay behaviour under
transient conditions. PSCAD/EMTDC does allow manufacturers to provide binary
library les of their custom models which can be used as a black-box, thus keeping
their design secret.
Even if all the design details were available, it would be an extremely complex
exercise to model all the electronic, electromechanical and software components of
the relay. A more practical approach is to develop models that match the behaviour
of the actual relays under specied operating conditions.
A practical way of developing suitable models is by means of comprehensive
validation, via physical testing of the actual relays and their intended models. The
Real-Time Playback (RTP) systemor Real-Time Digital Simulator (RTDS), described
in Chapter 13, provide the ideal tools for that purpose. The RTDS performs electro-
magnetic transient simulation to provide the fault current and voltage waveforms in
digital form. These are converted to analogue signals by means of A/D converters
and then amplied to the appropriate levels required by the relay.
Recognising the importance of relay modelling, a Working Group of the IEEE
Power System Relaying Committee has recently published a paper reviewing the
present state of the art in relay modelling and recommending guidelines for further
work [7].
8.4.1 Transducers
The performance of high-speed protection is closely related to the response of the
instrumentation transformers to the transient generated by the power system. There-
fore, to be effective, electromagnetic transient programs require adequate modelling
of the current transformers, magnetic voltage transformers and capacitor voltage
transformers. A Working Group (WG C-5) of the Systems Protection Subcommittee
206 Power systems electromagnetic transients simulation
of the IEEE Power System Relaying Committee has recently published [8] a com-
prehensive report on the physical elements of instrumentation transformers that are
important to the modelling of electromagnetic transients. The report contains valuable
information on the mathematical modelling of magnetic core transducers with specic
details of their implementation in all the main electromagnetic transient packages.
More detailed references on the specic models being discussed are [9][16].
CT modelling
The transient performance of current transformers (CTs) is inuenced by various
factors, especially the exponential decaying d.c. component of the primary current
following a disturbance. This component affects the build-up of the core ux causing
saturation which will introduce errors in the magnitude and phase angle of the gener-
ated signals. The core ux consists of an alternating and a unidirectional component
corresponding to the a.c. and d.c. content of the primary current. Also a high level of
remanence ux may be left in the core after the fault has been cleared. This ux may
either aid or oppose the build-up of core ux and could contribute to CT saturation
during subsequent faults, such as high-speed autoreclosing into a permanent fault,
depending on the relative polarities of the primary d.c. component and the remanent
ux. Moreover, after primary fault interruption, the CT can still produce a decaying
d.c. current due to the magnetic energy.
The EMTP and ATP programs contain two classes of non-linear models: one of
them explicitly denes the non-linearity as the full = f (i) function, whereas the
other denes it as a piecewise linear approximation. These programs support two
additional routines. One converts the r.m.s. vi saturation curve data into peak i
and the second adds representation of the hysteresis loop to the model.
In the EMTDC program the magnetising branch of the CT is represented as a
non-linear inductor in parallel with a non-linear resistor. This combination, shown
in the CT equivalent circuit of Figure 8.9, produces a smooth continuous BH loop
representation similar to that of EMTP/ATP models. Moreover, as the model uses
the piecewise representation, to avoid re-evaluation of the overall system conduction
matrix at any time when the solution calls for a change from one section to the next,
the non-linearity of both the inductive and resistive parallel branches are combined
into a voltage/current relationship. These voltage and current components cannot
L
p
L
s
R
p
R
s
ldeal
transformer
Non-linear
hysteretic
inductor
Non-linear
resistor
B
u
r
d
e
n
Figure 8.9 Detailed model of a current transformer
Control and protection 207
100
80
60
40
20
0
20
40
60
80
C
u
r
r
e
n
t

(
a
m
p
s
)
Time (ms)
25 50 75 100
80
60
40
20
0
20
40
60
80
C
u
r
r
e
n
t

(
a
m
p
s
)
Time (ms)
25 50 75 100
(a)
(b)
Figure 8.10 Comparison of EMTP simulation (solid line) and laboratory data (dot-
ted line) with high secondary burden. (a) without remanence in the CT;
(b) with remanence in the CT.
be calculated independently of each other, and an exact solution would require an
iterative solution. However, various techniques have been suggested to reduce the
simulation time in this respect [9].
Other important CT alternatives described by the Working Group document are
the Seetee [15] and the JilesAtherton [13] models. All these models have been tested
in the laboratory and shown to produce reasonable and practically identical results.
An example of the comparison between the EMTP simulation and laboratory data is
shown in Figure 8.10. These results must be interpreted with caution when trying to
duplicate them, because there is no information on the level of CT remanence, if any,
present in the laboratory tests.
CVT modelling
In the CVT the voltage transformation is achieved by a combination of a capacitive
divider, which achieves the main step-down in voltage, and a small wound voltage
208 Power systems electromagnetic transients simulation
C
2
C
1
C
t
C
ps
R
t
R
p
R
s
L
p
L
s
C
s
L
t
B
u
r
d
e
n
Ferroresonance
suppression
circuit
Figure 8.11 Detailed model of a capacitive voltage transformer
transformer. Adetailed model equivalent of the CVT is shown in Figure 8.11. Acom-
pensating reactor (normally placed on the primary side of the voltage transformer) is
used to minimise the equivalent source impedance at the fundamental frequency by
tuning it to the capacitance of the C
1
//C
2
combination; this reduces the fundamental
frequency voltage drop, which can otherwise cause a large error when the burden is an
electromechanical relay which draws a relatively high current. Figure 8.11 also shows
a ferroresonance suppression circuit to protect against a possible resonance between
the capacitors and a particular value of the combined inductance of the tuning non-
linear reactor and the magnetising inductance of the transformer. The incorporation
of winding capacitances C
t
and C
ps
increases the frequency range of the model (these
are not needed when the frequency range of interest is belowabout 500 Hz). Although
not shown in the gure, the CVTmay contain spark gaps, saturating inductors or metal
oxide varistors across the compensating inductor or ferroresonance suppression cir-
cuit and their operation during possible resonant conditions must be represented in
the model.
VT modelling
The modelling of magnetic voltage transformers is similar to that of other instrumen-
tation transformers. However, the large inductance of the primary winding and the
importance of heavy saturation and hysteresis loop require special attention.
8.4.2 Electromechanical relays
The constituent parts of electromechanical relays, i.e. electrical, mechanical and mag-
netic, can be separated when developing a mathematical model. Figure 8.12 shows a
diagram of the components of a model suitable for EMTP simulation [17], [18]. The
relay burden is represented as a function of frequency and magnitude of the input
current and includes the saturation non-linearities; this part is easily represented by
the electrical circuit components of the EMTP method. The mechanical and mag-
netic parts involve mass, spring and dashpot functions all of which are available in
the TACS section of the EMTP programs.
Control and protection 209
Relay
impedance
Mechanical
and magnetic
equations
Output
Input
Contacts
EMT power
system
representation
Relay model
to circuit breaker model
Figure 8.12 Diagram of relay model showing the combination of electrical, mag-
netic and mechanical parts
An alternative approach is to represent the dynamic behaviour of the relay by a
differential equation [19] of the form:
F = a x +b x +cx (8.20)
where
F is the difference between the applied and restrain forces
x is the distance travelled by the relay moving contact
a, b, c are empirically derived constants.
When the distance x is equal to the contact separation the relay operates.
8.4.3 Electronic relays
Electronic relays consist purely of static components, such as transistors, gates, ip-
ops, comparators, counters, level detectors, integrators, etc. and are considerably
more complex to model than electromechanical relays.
However, the TACS section of the EMTP can be used to represent all these compo-
nents; simple FORTRANstatements used for logical operations can also be modelled
in TACS. Abrief prefault simulation (say one or two cycles) is needed prior to transient
initialisation.
A detailed description of two specic distance protection relays is given in
references [20], [21].
8.4.4 Microprocessor-based relays
Digital relays normally use conventional distance measuring principles, such as the
phasor-based mho-circle. The required voltages and currents have to be sampled at
discrete points and the resulting information is used to derive their phase values. The
main components required to extract the fundamental frequency information are an
anti-aliasing input lter, an ADC(analogue to digital convertor) and a Fourier detector
as shown in the diagram of Figure 8.13.
210 Power systems electromagnetic transients simulation
EMTP-type
program
Relay model
Filter ADC Detector
v
,
i
v
,
i
v
,
i v
,
i
S
a
m
p
l
e
s
P
h
a
s
o
r
s
T
r
i
p
T
r
i
p
P
h
a
s
o
r
s
S
a
m
p
l
e
s
SLG
Figure 8.13 Main components of digital relay
The complete model would involve obtaining the circuit diagram and using the
EMTP method to represent all the individual components, i.e. resistors, capacitors,
inductors and operational ampliers. A more practical alternative is to obtain the
characteristics of the input lter, with the number of stages and signal level loss, etc.
With this information a reasonable model can be produced using the s-plane in the
TACS section of the EMTP program.
8.4.5 Circuit breakers
The simulation of transient phenomena caused, or affected, by circuit-breaker
operations involves two related issues. One is the representation of the non-linear
characteristics of the breaker, and the other the accurate placement of the switching
instants.
The electrical behaviour of the arc has been represented with different levels of
complexity, depending on the phenomena under investigation. In the simplest case
the circuit breaker is modelled as an ideal switch that operates when the current
changes sign (i.e. at the zero crossing); no attempt is made to represent the arc/system
interaction.
A more realistic approach is to model the arc as a time-varying resistance, the
prediction of which is based on the circuit-breaker characteristic, i.e. the effect of the
system on the arc must be pre-specied.
In the most accurate models the arc resistance dynamic variation is derived from
a differential or integral equation, e.g.
F =
_
t
2
t
1
(v(t ) v
0
(t ))
k
dt (8.21)
Control and protection 211
where v
0
and k are constants, and t
2
is the instant corresponding to voltage breakdown,
which occurs when the value of F reaches a user-dened value.
In the BPA version the voltagetime characteristic is simulated by an auxiliary
switch in which the breakdown is controlled by a ring signal received fromthe TACS
part of the EMTP.
The above considerations refer to circuit breaking. The modelling requirements
are different for the circuit-making action. In the latter case the main factor affecting
the transient overvoltage peak is the closing instant. Since that instant (which is
different in each phase) is not normally controllable, transient programs tend to use
statistical distributions of the switching overvoltages.
Considering the infrequent occurrence of power systemfaults, the switchings that
follow protection action add little overhead to the EMTP simulation process.
8.4.6 Surge arresters
Power system protection also includes insulation coordination, mostly carried out by
means of surge arresters [22].
Most arresters in present use are of the silicon carbide and metal oxide types. The
former type uses a silicon carbide resistor in series with a spark gap. When the over-
voltage exceeds the spark-over level (Figure 8.14) the spark gap connects the arrester
to the network; the resistor, which has a non-linear voltage/current characteristic (such
as shown in Figure 8.15) then limits the current through the arrester.
Time
0
0
Incoming
wave
Breakdown
occurs
V
Breakdown
V
o
l
t
a
g
e
Figure 8.14 Voltagetime characteristic of a gap

Potrebbero piacerti anche