Sei sulla pagina 1di 7

Brushless excitation of synchronous generators:

study of models and control optimization


Nicolau, Nuno, IST

 systems have an almost instantaneous response. DC excitation


Abstract— Brushless excitation systems have been largely systems have been less used in favor of AC systems that avoid
applied in recent years. Nonetheless models used to simulate bushes and rings.
them show a few problems that originate imprecise results. This Excitation systems are, normally, composed of: Automatic
work has the objective to program and simulate the AC7B model
voltage regulator (AVR), an exciter, measuring elements, a
for brushless excitation systems. This model is an update on the
AC1A model from IEEE. The simulation will, first, compare the power system stabilizer (PSS), and a limitation and protection
performances of both models in a short-circuit fault situation. unit, as shown in figure 1.
After that new values for the AC7B’s controller will be calculated
in order to improve its response to the short-circuit fault.

Index Terms—Excitation systems, AC1A, AC7B, PID


controller

I. INTRODUCTION

S synchrouns generator are, today, the base element of


Electric Energy Systems [1]. These generators have a
excitation system that produces a excitation current necessary
for the generator to work. Moreover, excitation systems can Fig. 1. Excitation system.
control the generators voltage, making them an important
element in system stability. To ensure that generators operate The models are usually divided in three groups: static
rightly it is crucial to understand excitation systems. For that systems (ST); dc systems (DC); and ac systems (AC).
simulation is essential. Today must of the operation and
control of Electric Energy Systems is performed through A. Static Systems
simulation. Excitation systems are modeled and simulates with
Every component of this excitation system is static. The
computational models. Those models have been used for a
current is provided to the generator through transformers or
number of years. In 1968, IEEE published a series of models
auxiliary generators winding. The rectifiers use brushes and
for excitation systems. These models were updated in 1981,
deliver the current directly on the generator’s field winding.
and became industry standards for the test of excitation
By acting directly on the generator’s field winding the
systems [2]. Still, some problems are encountered when using
response time is shortened. The current is controlled by an
these models. Brushless excitation systems are some of the
AVR. Some of the disadvantages of these systems are the use
most used and worse simulated. It becomes then necessary to
of brushes and the deformations in the excitation current
develop models that allow for a better simulation of excitation
caused by the semiconductors. Despite this, the response of
systems.
static systems is practically instantaneous and these systems
are less costly, and because of that they have a satisfactory
performance for generators in big networks.
II. EXCITATION SYSTEMS
As mentioned before, excitation systems are responsible for
the excitation current of generators, and are able to control its
terminal voltage. The basic requirement for excitation systems
is to provide and adjust the excitation current of generators to
maintain the terminal voltage in the right levels. Over the
years, different types of excitation systems have been applied.
The first ones had to be adjusted manually. Automatic
regulators started to gain visibility, and today excitation
Fig. 4. Static AC excitation system.
Fig. 2. Static excitation system.
Systems with rotary rectification eliminate brushes and
rings. They’re used for big generator where the field winding
B. DC Systems can reach MW. An excitation system like this has the
For DC excitation systems dc generators are used as the following elements:
source for the excitation, which is delivered in the rotor of the
main generator using rings. If the dc generator is excited using 1) Inside-out auxiliary ac generator.
an external source, this is made with a permanent magnet 2) Non-controlled rectifier.
generator (PMG). These systems were highly used between 3) An exciter.
1920 e 1960 [3]. In the mid 1960 dc systems usage slows 4) Controlled rectifier.
down. DC systems show some disadvantages with the increase 5) An AVR.
of generators nominal power. The great number of brushes
necessary for high currents e low voltages is one of them. A brushless excitation system is, essentially, an inside-out
ac generator that delivers its ac voltage to the rotor of the main
generator, and receives its excitation from the stator of that
same generator. The ac voltage of the inside-out generator is
transformed to dc by a non-controlled rectifier, this way
producing the excitation current of the main generator. An
exciter, in the stator of the main generator, produces a dc
current that controls the output voltage of the inside-out
generator. The dc current of the exciter comes from a
controlled rectifier, either manually or by an AVR. The exciter
can be powered by a PMG, this way avoiding any external
voltage source, figure 5.

Fig. 3. DC excitation system.

C. AC Systems
In these systems auxiliary ac generators, mounted on the
main generator’s shaft, are used as the source for the
excitation current. The output of the auxiliary generator is
rectified by rectifiers, that can be controlled or non-controlled,
that produced the current necessary for the main generator.
The rectifiers can be stationary or rotary.
The AC systems with stationary rectifiers need rings to Fig. 5. Rotary AC excitation system.
deliver the current to the main generator, figure 4. The use of
rings is a major disadvantage that led to the search for new
alternatives to eliminate them. Some of the advantages of brushless excitation system are:
easy manual control; reduced maintenance.

III. AC1A MODEL


The AC1A model from IEEE represents a brushless
excitation system, with non-controlled rectifiers, figure 6.
This model is widely used today to simulate brushless
excitation systems. However the model leads to some
mistakes, and doesn’t accurately represent the excitation
system. One of the bigger problems with this model is that it
doesn’t assume the influence of the terminal voltage of the
main generator.

Fig. 6. AC1A model.


IV. AC7B MODEL
The model is linear with the exception of the exciter
saturation function, SE[VE], and the rectifier regulation A. The model
function, FEX. The AC7B model, figure 8, is an updated version of some
The SE[VE] function can be representing by the following of the previous excitation systems, mainly the AC1A. Some of
expression: its features are an internal feedback loop, which regulates the
filed voltage, a faster excitation current limit, and a PID
𝑆𝐸 (𝑉𝐸 ) = 𝑎 × 𝑒 𝑏∗𝑉𝐸 (1) regulator.

The a and b variables are calculate from two points of the


saturation function, (S1, E1) e (S2, E2), with:

𝐸 ×𝑆
𝑙𝑛(𝐸1 ×𝑆1 )
2 2
𝑏= (2)
𝐸1 −𝐸2
𝐸1 ×𝑆1
𝑎= (3)
𝑒 𝑏×𝐸1

The rectifier regulation function, FEX, receives the value of


IN and, based on that, determines one mode of operation. Fig. 8. AC7B model.
There exist here modes of operation, figure 7.

B. Programming the model in Matlab


The main objective of this work was to program the AC7B
model in Matlab, with the Simulink environment. This was
made like figure 9 shows. The model has 5 inputs, VC, VREF,
VS, VUEL, and IFD and one output, EFD.

Fig. 7. Rectifier regulation curve.

The curve is described in segments through the next


expressions:

𝐹𝐸𝑋 = 1 𝐼𝑁 ≤ 0
Fig. 9. Programmed model.
𝐹𝐸𝑋 = 1 − 0.577 × 𝐼𝑁 0 < 𝐼𝑁 ≤ 0.433
𝐹𝐸𝑋 = √0.75 − 𝐼 2 𝑁 0.433 < 𝐼𝑁 < 0.75 (4)
𝐹𝐸𝑋 = 1.732(1 − 𝐼𝑁 ) 0.75 ≤ 𝐼𝑁 ≤ 1
𝐹𝐸𝑋 = 0 𝐼𝑁 > 1
V. SIMULATION OF THE AC1A AND AC7B MODELS
TABLE II
GENERATOR PARAMETERS
A. Simulation environment
After the model is programmed, it’s important to understand Parameter Value
how it works. For that, the model was simulated with a PN 12.153E6
generator and a load for a short-circuit test, figure 10. It this VN 11E3
it’s possible to observe the response of the system to a fn 50
Xd 1.65
situation that causes a lot of disturbances. The excitation Xd’ 0.14
system must be able to limit the overvoltage and guarantee Xd’’ 0.12
that the voltage goes back to its initial value. Xq 1.5
Xq’’ 0.23
Xl 0.119
Tdo’ 3.2
Td’’ 0.045
Td’ 0.245
Tdo’’ 0.45
Tq’’ 0.067
Tqo’’ 0.438
Rs 0.0038
H 6
F 0.001
Pole pairs 2
Pole type Salient

Fig. 10. Simulation environment. Using the values on the tables, both models were simulated
. Figures 11, 12, and 13 show the responses for the short-
Both models were tested for a short-circuit fault of 1 circuit fault.
second, with a voltage fall of 20, 50, and 80%. Tables I and II
show the parameters used for each model, and for the
generator. All of these values were obtained from a real
system.
TABLE I
MODEL PARAMETERS

Parameter AC1A AC7B

TR 0 0.02
KPR - 36.57
KIR - 73.14
KDR - 0
TDR - 9999
TC 25 -
TB 1 -
KA 400 -
KP - 0 Fig. 11. Models response to a short-circuit, with a 20% voltage fall.
KPA - 8.07
KIA - 40.33
TA 0.03 -
KF 0.0000001 -
KF1 - 0
KF2 - 1
KF3 - 0
TF 1 9999
KI 0 -
KL - 10
VAMAX 14.5 47.19
VAMIN -14.5 - 47.19
VRMAX 6.03 34.45
VRMIN -5.43 0
VUEL -15 -
VOEL 15 -
VFEMAX - 34.45
VEMIN - 0 Fig. 12. Models response to a short-circuit, with 50% voltage fall.
TE 0.15 0.55
KE 1 1
SEVE1 0.1 1.70
SEVE2 0.01 1.14
VE1 4.18 12.76
VE2 3.14 11.36
KD 0.38 2.44
KC 0.2
Fig. 13. Models response to a short-circuit, with 80% voltage fall.

Fig. 15. AC7B responses for two values for the upper regulator limit.
From the figures we can see that both of the models have a
difficulty in limiting the overvoltage and have a slow
response. After the fault the voltage shouldn’t surpass 10% of By changing VRMAX its possible to limit the overvoltage
the nominal value, although in some cases it can go as high as after the fault. Still the response of the system is a bit slower.
40% for a few seconds. Either way, both of the models have To improve this we can find new values for the regulator,
values higher than 40% when the voltage fall is around 80%. instead of just for the upper limit.
This situation is in need of improvement.
One easy solution is to decrease the voltage regulator upper
limit to a value that sets the overvoltage to 10%. In both VI. REGULATIONS OPTIMIZATION OF THE AC7B’S CONTROLLER
models this leads to a situation where the response becomes
very slow. This happens because of the windup in the In order to find new values for the controller two methods
regulator. In the AC7B model it is possible to introduce anti- were used. First the PID Tuner tool in Matlab. The tool
windup in its PID regulator to fix this problem. The AC1A provides a simple way to determinate values for a PID
model doesn’t allow this solution. Therefore the next controller. Second the Ziegler-Nichols method. This method
simulations will only consider the AC7B model. Figure 14 provides a formula to calculate the PID parameters.
shows the models response to the same situation as figure 13,
with anti-windup.
A. PID Tuner
Before using this tool the system need to be simplified. For
that, the generator has to be replace with a bloc that simulates
its behavior, through the transfer function: KG/(1 + s.TG). For
this optimization method, we consider the saturated values of
the generator’s parameters, and so KG is 1 and TG equals
4,591. The AC7B model has its regulator replaced with a PID
Controller bloc, available in Matlab, that allows for the use of
the PID Tuner tool. Figure 16 shows the system with these
changes.

Fig. 14. AC7B response to a short-circuit, with 80% voltage fall and anti-
windup.

With the introduction of anti-windup the model’s response


becomes slower. As said earlier, in some cases the overvoltage
limit can be higher, about 40%, as long as it stays higher than
10% for less than 5 seconds. This can be used to try and get a
faster response, altering again the upper limit of the voltage Fig. 16. Simplified system used for the PID Tuner tool.
regulator, this time with a value that allows the overvoltage to
reach 40%. Figure 15 shows the responses of the AC7B model
with different values for the upper regulator limit.
B. Ziegler-Nichols
Opening the PID Tuner tool shows the response of the The Ziegler-Nichols method was developed by John G.
system with the original values and the response with the new Ziegler and Nathaniel B. Nichols. With a few steps it’s
values, figure 17. The new values are calculated by the tool. possible to determine two values, the ultimate gain, Ku, and
By changing the arrow at the bottom it’s possible to see the the oscillation period, T u, with which the PID parameters are
new response changing, as well as the tuned values. calculated.
First, the integral gain, KI, and the derivative gain, KD, are
set to zero. Then the proportional gain, KP, is increased until
the system start to oscillate with a constant amplitude, figure
19.

Fig. 17. PID Tuner.

The values obtained with the tool were then tested with the
complete system. The new values for the PID controller found
with the PID Tuner were:
Fig. 19. Constant oscillation with Tu period.
 KPR = 49.1719
 KIR = 26.1069 With the values of Ku and Tu the PID parameters are
 KDR = 2.6773 calculated like:

 KP = Ku/1,7
Figure 18 show the response to the short-circuit fault, with  KI = 2*(KP/Tu)
80& voltage fall, with the new and the old values for the PID  KD = (KP*Tu)/8
controller parameters.

Using the Ziegler-Nichols method for this case, it was


obtained a Ku and Tu of 26 and 1,8, respectively. With this
values the we have:

 KP = 15.29
 KI = 28.89
 KD = 5.85

With this values the AC7B system was simulated, this time
for a short-circuit fault that lowers the voltage to zero, figure
Fig. 18. AC7B responses with the old and the new values for PID 20. For the same fault situation, voltage dropping to zero,
controller parameters. figure 21 shows the response with the values calculated with
the PID tuner tool.
1.8
REFERENCES
1.6

1.4
[1] Kundur, P. “Power System Stability and Control”, McGraw-Hill,
1.2 1994.
Vt (pu)

1
[2] Sucena Paiva, José Pedro “Redes de Energia Eléctrica : uma Análise
0.8 Sistémica”, IST Press 2005.

0.6
[3] IEEE Power Engineering Society, “IEEE Recommended Practice for
0.4 Excitation System Models for Power System Stability Studies”, IEEE
Std. 421.5-2005, 2006
0.2

0
34 35 36 37 38 39 40 41
[4] Kabir, S.M.L. and Shutleworth, R. “Brushless exciter models”, IEEE
Tempo (s) Proc.-Gener. Transm. Dirtib., Vol. 141, No. 1, January 1994.
Fig. 20. AC7B response with PID controller parameters from Ziegler-
Nichols method. [5] Jerkovic, V., Miklosevic, K. e Spoljaric, Z. “Excitation System
Models of Synchronous Generators”.

[6] There, M., Chawardol, P. e Badre, D. “Excitation System of


Alternator”, International Journal of Engineering Research &
1.8
Technology (IJERT), Vol. 2 Issue 2, February 2013.

1.6
[7] Feng, S., Jianbo, X., Guoping, W. e Yong-hong, X. “Study of a
1.4 Brushless Excitation System Parameters Estimation Based on
Improved Genetic Algorithm”, IEEE,
1.2

[8] Bayram, M., Bulbul, H., Can, C., Bayindir, R. “Matlab/GUI Based
Vt (pu)

1
Basic Design Principles of PID Controller in AVR”, Istanbul, Turkey,
0.8
13-17 May 2013.
0.6

0.4

0.2

0
34 35 36 37 38 39 40 41
Tempo (s)

Fig. 20. AC7B response with PID controller parameters from PID Tuner.

VII. CONCLUSION

In conclusion it can be said that the study of brushless


excitation systems is still a topic that needs study, given the
crucial importance these systems have in electric energy
systems.
As far as this particular work goes, the objective of building
the AC7B model in Matlab was achieved, and with this it was
possible to understand the model’s behavior in a situation of a
short-circuit fault. When compared with the AC1A model, the
AC7B had a faster response. The improvements that the
AC7B model has allow for a better simulation of the brushless
excitation systems. The PID regulator, in particular, can be
used to regulate the systems response better. A number of
methods can be used for that purpose. Using the Ziegler-
Nichols method was simple, but it provided a response far
from optimal. The PID Tuner tool was not as direct but it yield
values for the PID parameters that improved the original
response of the system.
It is also important to point out that, as a future work, it
would be relevant to have a build in bloc in Matlab for the
AC7B model.

Potrebbero piacerti anche