Title: Automatic Voltage Regulator and Power

System Stabilizer Type Tests for Russian

Certification
Paper ID No: 20
Technical Track
Session 3: Power System Stabilisation via
Asset Optimisation
Author: Sandro Tuscano
Company: Ansaldo Energia
Country: Italy
Abstract
Ansaldo Energia is completing in Russia the supply of 14 turbogenerator sets equipped
with gas turbine AE64.3A. One of the related activities was to obtain the System
Operator certification necessary for installation of the AVR (Automatic Voltage
Regulator) in the Russian Power System. The AVR and its embedded PSS (Power
System Stabilizer) were tested in multiple sessions at the NIIPT research center by
connecting them to the Electro-Dynamic Simulator EDS, a physical model including
reduced power generators (around 10 kVA), transformers, loads, lines and the related
control and recording equipment.
Despite the fact that the regulator under test has already been installed in about 30
countries, during the NIIPT tests several parts of the regulation program had to be
modified. In fact new functions had to be added which, although not requested by
international standards nor by national grid codes, are specifically required for energy
production in the Russian Federation, in order to increase the stability of the Power
System and to prevent disconnection of the generating units.
During the NIIPT tests, the AVR underwent dozens of short circuits, fast frequency
variations, direct coupling between generators and other tests that cannot be
accomplished in real power plants, where the grid cannot be pushed close to its stability
limits.
After the above mentioned modifications, the AVR successfully passed the official test
program, composed of more than 80 tests specifically devised to guarantee a good
behaviour in all critical transients, as damping local and interarea oscillations, field
voltage forcing in case of short circuit, or correct action of underexcitation and rotor
current limits.
It is hard to obtain such accurate and reliable results with a computer simulation,
because of the differences between the software program, the microprocessor and input-
output hardware, and the noise that affects the real trends.

Introduction
In recent years Ansaldo Energia cooperated with Russian customers as Power
Machines, Stroytransgaz, Enel OGK-5 and Mosenergo, in the building of several power
plants located in the Russion Federation, for example at Saint Petersbourg, Tyumen,
Adler and Sredneuralskaya. Eleven of these power plants has already been
commissioned, and others are under construction. In these projects Ansaldo supplies the
turbine-generator set outfitted with the turbine control system, the protection system and
the excitation system. Most of these units are equipped with the AE64.3A gas turbine,
connected to a 82 MVA generator; one of them is equipped with the AE94.3A gas
turbine associated with a 300 MVA generator, in combined cycle with an RT20 steam
turbine connected to a 185 MVA generator.
The Russian Federation has a particular Power System that can operate close to the
stability limits of the grid, especially in case weak transmission lines have to transfer
heavy loads. In these conditions the voltage regulators must be capable of a strong
damping action, otherwise the stability could be affected by increasing spontaneous
oscillations. Also for this reason the Russian requirements are more severe than other
countries, implying that, before a generating unit can be connected to the Russian
system, special tests must be carried out to verify the voltage regulator performances
and appropriate certification has to be obtained.
Such tests are performed by the power transmission research institute NIIPT of Saint
Petersburg, where the largest existing physical model for electro-dynamic simulations is
located. This model makes use of about 70 generators, 150 transformers, 160 loads, 700
transmission lines and a control room equipped with the instrumentation for command,
measurement and monitoring, including a 32 channels digital recorder. The generators
size ranges between 1,5 and 30 kVA and is designed to endure overcurrents and
overvoltages without damage, even when the limiting functions of the regulator under
test are not working properly; some of the generators are equipped with special devices
to modify the inertia and the time constant.

Test program
To verify the behavior of the voltage regulator, the NIIPT research institute has
developed a protocol envisaging 81 tests in different grid configurations and operating
conditions, including:
• Single phase and phase-to-phase short circuits
• Set-point reference steps
• Increase of power transmission and low frequency oscillations
• Grid transients with line openings
• Increase and reduction of the grid voltage
• Direct connections between generators terminals, without interposed transformers
These events are used to verify the regulator efficiency, and in particular:
• The PSS capability to damp down oscillations
• The fast field forcing in case of stator voltage drop
• The correct action of the limiting functions
• The correct dynamic response during transients
• The regulators capability to prevent reactive power circulation between their
generators

Due to the last group of tests, a couple of twin regulators has to be installed in the
laboratory grid.

The voltage regulators installed in the above mentioned power plants are manufactured
by EAA in the north of Italy, at Lerma. The regulator has a main loop for stator voltage
regulation and an inner loop for excitation current regulation (Figure 1).
 REACTIVE POWER STATOR VOLTAGE FIELD CURRENT
REFERENCE REFERENCE REFERENCE

REACTIVE POWER STATOR VOLTAGE INTERNAL LOOP

REGULATOR REGULATOR REGULATOR
+25% Va_spmax Iemax Vcp

COSΦ P cosφ + 1+s TR + + 1+s T3 + 1+s T2 FIELD

KR K3A K2A KCEA
REFERENCE sinφ s s s VOLTAGE
- + - -

+
-25% Va_spmin Iemin Vcn
ACTIVE
POWER (P)
STATOR FIELD
VOLTAGE CURRENT
COMPOUND
+25% PSS_lim PSS (POWER SYSTEM STABILIZER)

REACTIVE KCR + 1+s TL3 1+s TL1 s TW2 s TW - ACTIVE

K KP
POWER 1+s TC 1+s TL4 1+s TL2 1+s TW2 1+s TW POWER
+ +
-25% -PSS_lim
KF FREQUENCY

UNDEREXCITATION
LIMIT FLUX LIMIT
+35% 0%
MINIMUM -
+ 1 + + 1 +
REACTIVE KFL FREQUENCY
s Tu s TFL
POWER -
0% -20% STATOR VOLTAGE

ROTOR CURRENT
LIMIT
0%
MAXIMUM + 1
FIELD
s TEL
CURRENT -
-35%
FIELD
CURRENT

Figure 1 AVR block diagram

This model of regulator was developed in 1998/1999, and has been used with more than
150 turbogenerators installed by Ansaldo Energia in about 30 different countries.
Throughout the last years, the regulator was outfitted with further functions so as to
comply with the requirements of different National Grid Codes and customer
specifications.
Despite this extensive trial run, the regulator was still not conforming to Russian
requirements in some specific conditions, as emerged during the 2011 testing program
at the NIIPT laboratory.
During the test sessions in Saint Petersburg, the following main questions have been
considered and consequent improvements have been achieved:
1) Effectiveness of PSS damping action
2) Pss inhibition in case of large frequency variations
3) Relè forcing during short circuit
4) Rotor current limiting

1) Power system stabilizer (PSS)

The PSS tests in the NIIPT program have been divided in two different groups to allow
verifying the following:
a) Dynamic local machine oscillation tests, which verifies the PSS damping capability
of oscillations in the range between 0,8 and 2 Hz.
b) Static aperiodic stability tests, to check the behavior at lower frequencies, ranging
between 0,1 and 0,8 Hz.

Initially, the PSS scheme shown in Figure 2 went through the stability tests, with 2
inputs (active power and frequency), 2 wash-out blocks and 2 lead-lag blocks. The
frequency is taken from the generator terminal voltage signal, without the need of a
speed feedback .
 PSS (POWER SYSTEM STABILIZER) PSS_lim

ACTIVE - s TW s TW2 1+s TL1 1+s TL3

KP K
POWER 1+s TW 1+s TW2 1+s TL2 1+s TL4
+
-PSS_lim
FREQUENCY KF

Figure 2 Initial PSS scheme.

In the first tests the PSS parameters were tuned with a simplified method, and thus not
optimized for the network of the test circuit. Consequently the results of tests of group
b) were not satisfactory, due to the presence of a 0,24 Hz oscillation that was not
damped by the PSS, as show in Figure 3.

Figure 3 Test 19. Active power Pr1 transmission on line 1-3 increase, for aperiodic
stability. Line 4-2 is disconnected. PSS is active. Test failed.
U_1b = voltage at bus bar 1
P_Г-1 = generator active power
P_1-3 = active power in line 1-3

The PSS tests of group b), relating the aperiodic stability, are particularly tough for the
regulator, since the increased active power transmission in the line leads to instability,
and the other 3 generators of the test circuit do not contribute to the damping action
because their regulators has no PSS. The test circuit is shown in Figure 4.
 Ug=500 kV
Scheme 2. Mode 4 DUT
~ Рg=2000 МW

Р=1000 МW Р=1000 МW

2 3

Ug=503 kV Ug=503 kV
Рg=7200 MW Рg=2000 МW

~ 4
~
Рn=7200 МW Рn=2000 МW

Uг=476 kV

~
Х1-2=100 Оhm, Х1-3=100 Оhm,
Х2-4=Х3-4=50 Оhm

Figure 4. Test circuit for aperiodic stability. In the physical model the
voltage and power values are scaled.
The device under test (DUT) is connected to upper generator.

After this first result, a study was done to obtain an optimized tuning of the PSS
parameters, in combination with the tuning of the regulator. In fact, the PSS output adds
to the generator stator voltage reference, therefore the regulator transfer function in the
frequency range of interest must also be considered.
As the goal is to dampen the inter-area oscillations at low frequency, the more
efficacious action is obtained by the frequency channel of the PSS, which gain must be
tuned at a high value.
With the new set of parameters the stability in the inter-area test was reached, but the
new tuning caused a very nervous excitation voltage regulation, with an unacceptable
continuous correction of the field voltage that swung more than 40 % of the rated value.
Not to mention that an extreme tuning of some parameter specifically aimed to pass one
single test, could cause problems in the other tests, because the complete qualification
program must be passed with the same set of regulator and PSS parameters, so the
tuning must be balanced for all situations.

In order to reach a radical and fresh solution to this problem the PSS block diagram was
completely changed, moving to a more efficient configuration. Two ways were
possible:
• Migration to the IEEE PSS2B model
• Adoption of the model shown in Figure 5, already widely and successfully used by
Russian manufacturers
Owing to the tight cooperation with Power Machines and Electrosila, the second
solution was chosen. For the first time, this scheme was used with a 2 loop regulator,
contrary to previous applications on regulators without an internal excitation current
loop.

Figure 5 Modified PSS scheme.

The 4 inputs of this type of PSS are:
- frequency deviation
- frequency derivative
- rotor current derivative
- stator voltage derivative

The switch of the PSS scheme to the Russian model allowed to obtain a good result in
the aperiodic stability test (Figure 6),as well as in all the other tests of the program.
Figure 6 Test 19. Active power transmission on line 1-2 increase, for aperiodic stability.
Line 3-4 is disconnected. PSS is active. Test successful.
U_1b = bus bar 1 voltage
P_Г-2 = active power of the generator controlled by regulator under test
P_1-2 = active power to line 1-2

Figure 7 shows the difference in case the PSS was switched off, resulting in active
power oscillations in the transmission line 1-2.

Figure 7 Test 19 and 20. Aperiodic stability test, comparison between PSS on (red trace)
and PSS off (green trace)
P_1-2(I) = active power in line 1-2 in case of PSS active
P_1-2(II) = active power in line 1-2 in case of PSS not active

The main action of this type of PSS is given by the two frequency channels, while the
field current and stator voltage channels allow to expand the stability area.
By a selection parameter it is possible to set one of the following configurations:
a) the PSS output is added to the stator voltage reference
b) the PSS output is added to the excitation current reference
The best result has been obtained with scheme b), corresponding to Figure 8 with switch
PSS-SW closed and switch PSS1-SW opened.

FIELD CURRENT
REFERENCE

STATOR VOLTAGE INTERNAL LOOP

REGULATOR REGULATOR
Va_spmax Iemax Vcp

+ 1+s T3 + 1+s T2 FIELD

K3A K2A KCEA
s s VOLTAGE
- -
+

+
Va_spmin Iemin Vcn
STATOR FIELD
PSS1 VOLTAGE PSS CURRENT
SW SW

FREQUENCY
PSS ROTOR CURRENT
STATOR VOLTAGE

Figure 8 . PSS insertion in the regulation loop.

The critical issue of this PSS scheme is the quality of the frequency signal in terms of
resolution and cleanness. Initially the frequency was acquired by an analog signal
generated inside the card, but this had a resolution of 16.6 mHz which proved to be
insufficient for the proper operation of the PSS.
Therefore the frequency acquisition was improved by using a digital counter at 1 MHz,
that allowed to reach the following accuracy in the operating area around 50 Hz:

The frequency acquisition is refreshed each 10 ms and the length of the period is
measured between two subsequent rising edges and, alternately, between two
subsequent falling edges.

2) PSS inhibition in case of large frequency variations

In case of unidirectional variations of frequency exceeding a certain gradient (> 50
mHz/s), that can be caused for example by disconnection of loads or groups of
generation, the PSS contribution may be counterproductive and may affect the stability
of the power system. This problem can be avoided by adding a special logic in the
AVR, and one of the NIIPT tests is specifically conceived to check this issue.
As the various grid codes and the buyer specifications don’t prescribe this corrective
tool, the regulator used by Ansaldo wasn’t provided with such a function. So it was
necessary to add a new logic to disable the PSS in the above described situation and
avoid detrimental variations of the stator voltage.
 UNDESIRED
GENERATOR
VOLTAGE
REDUCTION

Figure 9 Test 41 and 42. Frequency fall.

U_г-2(I)= generator voltage in case of PSS switched on, with inhibition
U_г-2(II) = generator voltage in case of PSS switched on, without inhibition
U_г-2(III)= generator voltage in case of PSS switched off
df = frequency deviation

UNDESIRED
GENERATOR
VOLTAGE
RISE

Figure 10 Test 37 and 38. Frequency rising.

U_г-2(I)= generator voltage in case of PSS switched on, with inhibition
U_г-2(II) = generator voltage in case of PSS switched on, without inhibition
U_г-2(III)= generator voltage in case of PSS switched off
df = frequency deviation

The execution of this function is delicate, because in order to be effective the

intervention must be very fast, within 300 ms, so the logic must elaborate the frequency
feedback and command the correct action in a very short time.
This task becomes more complicated because in some cases the PSS must not be de-
activated, even if the frequency derivative exceeds 50 mHz/s. This is the typical case of
a short circuit, after which the PSS must be working to damp down the oscillations.
As a consequence the regulator has been provided with a logic that discerns the events
characterized by unidirectional frequency variations from the events characterized by
frequency variations of oscillatory nature, as the transient following a short-circuit. In
fact in the first case the PSS must be blocked in order to avoid unjustified alterations of
the voltage, while in the second case the PSS must be free to operate.
The algorithm is based on the weighted average of the frequency derivative and the
frequency deviation. The resulting signal is then filtered and compared to a threshold
with an hysteresis.
The first release of the modified program was tested in Italy at the factory, where it
worked correctly, but the behavior in the NIIPT laboratory was unexpectedly different,
because of the much higher frequency short term variations and noise in the physical
model grid, causing several untimely activations of the function. Based on the NIIPT
tests, the function was calibrated in the final arrangement, successfully preventing
untimely interventions.

3) Relè forcing
Following a short circuit, the generator terminal voltage falls to a lower value. In case of
voltage drops under 80% of the pre-fault condition value, the Russian prescriptions
require a very fast field voltage forcing, that must reach the ceiling in less than 60 ms.
In normal operation and small transients the forcing to the ceiling value can be easily
accomplished within this time, or even less, but in case of a short circuit the stator
current has a sudden increase (typically up to 2-4 p.u.), generating an inducted rise of
the excitation current, which causes a transient de-excitation action by the inner field
current regulation loop (see Figure 11).

TRANSIENT
DE-EXCITATION
BEFORE FORCING

Figure 11 Test 44. Double phase short circuit. Test failed.

U_г-1 = generator voltage
if_г-1 = excitation current
Uf_г-1 = excitation voltage

It was possible to mitigate this undesirable effect by a different tuning of the regulator
parameters, but the result was still not enough fast, so it was necessary to add the
special ‘relé forcing’ function, that is already known and used in the Russian
applications.
The relé forcing moves back the thyristor firing angle to its minimum value, thus
forcing the maximum possible excitation voltage, when the generator voltage is
detected to drop under a threshold, typically set at 0,85 pu.
In order to obtain a more effective result, the software that triggers the relé forcing
checks not only the stator voltage value, but also the excitation current fast variation
which occurs in case of short circuit. The excitation current offers the benefit of a
reduced filter in the feedback, providing a fast action. The activation of the relé forcing
is given by the formula:

where:
stator voltage [p.u.], typically 0.85
Field current [p.u.]
time
stator voltage threshold [p.u.]
adjustable coefficient [p.u.], typical values between 0,1 and 0,6

Only as an indication, the virtual excitation current variation that would had caused the
relé forcing activation in case of rated stator voltage, with K=0,6, is given by:

This value is much higher than the maximum gradient that can be reached in the field
forcing generated by the regulator.
The exit from the relé forcing was accurately accomplished taking care of the filters and
integrators initialization, to avoid inappropriate transients that could reduce the PSS
effectiveness.
After the above mentioned improvements, the regulator was able to force a fast
excitation ceiling voltage, as shown in Figure 12.
 FORCING
ACTION

Figure 12 Test 44. Double phase short circuit. Test successful.

Uf_г-2 = excitation voltage
U_г-2 = generator voltage
if_г-2 = excitation current

This test was also useful to verify the correct action of the PSS after the short circuit.
The result of the same event carried out with the PSS disabled is illustrated in Figure 13.

UNDAMPED
OSCILLATION

Figure 13 Test 45. Double phase short circuit with PSS switched off.
Uf_г-2 = excitation voltage
U_г-2 = generator voltage
if_г-2 = excitation current
4) Limiters
A minor adjustment in the regulator was done also as concerns the rotor current limit
function.
During a short circuit, the rotor current is not controlled only by the regulation, but, due
to the induction from the stator current, rises up to values higher than the threshold of
the limit. This is emphasized by the oscillations of the rotor current at 50 and 100 Hz.
The logic of the limit intervention was modified: after a short circuit, the relé forcing
initially prevails over the rotor limit, and only after a while (about 150 ms) the limiter is
allowed to lower the excitation voltage, if necessary to bring back the excitation current
under the limit value. The effect of this function can be seen in Figure 13.

LIMITING ACTION

Figure 14 Test 56. Double phase short circuit of 0,4 s duration and rotor limiter
intervention.
Uf_г-2 = excitation voltage
U_г-2 = generator voltage
if_г-2 = excitation current

General remarks
The test campaign was useful to reveal some specific issues which cannot be detected
by other test methods, as computer simulations and site tests.
It is easy to understand that certain transients, as short circuits and frequency collapses,
cannot be carried out during site tests at the power plant. Moreover, the low frequency
inter-area oscillations are not often present during the site test, but they only emerge in
particular operating conditions.
On the other hand Ansaldo Energia developed a computer model to reproduce the AVR
block diagram and functions, the generator (with a fifth order model), the speed
regulator and the simplified grid, in Matlab/Simulink simulation platform. This tool is
certainly useful to check the theoretical behavior of the regulator, and gives a good
result in many situations, but it is not suitable to obtain a realistic result in some
particular case. In fact there is an approximation in the mathematical modeling of the
physical components, that is also caused by nonlinearities and construction tolerances;
moreover the same transfer functions of the block diagram are carried out in a different
way in the real AVR and in the computer simulation, because of the differences in the
software program that is written with different computing language and command lines;
and other differences are in the input-output hardware, in the sampling time, in the
integration method, in the noise that affects the real trends.

For example, to get a good result with issue n. 1 and 2 (PSS and PSS inhibition in case
of high frequency variations) the frequency acquisition must have a high resolution, and
its derivative must be calculated with a suitable method to obtain a clean signal without
an excessive delay. During the NIIPT tests, the method of derivation of the frequency
and other quantities, as stator voltage and excitation current, was improved by the use of
the following formula calculated in 2i +1 points:
x' = (i × (xi - x-i) + (i – 1) × (xi-1 - x-i+1) + ... + (x1 - x-1)) / (k × h)
where
k = (i × (i + 1) × (2i + 1)) / 3,
xi – result of the last measurement
i = 2 or 3, depending on the quantity concerned
h = 10 ms, sampling time
The derivative is calculated in the middle of 2i +1 points (e.g. 5 points if i=2), thus
introducing a delay T = i × h.

Conclusions
After all the necessary modifications and upgrades, the AVR successfully passed the
official test program and was recommended by the NIIPT Research Institute for
installation in the Russian Federation.
In the meanwhile, other versions of the regulation program ware developed, with the
following alternative regulation schemes:
a) Double loop regulation with inner excitation current loop (same as the tested
version) in combination with PSS IEEE 2B model, on the same hardware of the
tested AVR.
b) Single stator voltage regulation loop in combination with PSS IEEE 2B model, on a
new electronic card compatible with the actual AVR assembly
c) Single stator voltage regulation loop with PSS IEEE 2B model, on a new electronic
card and new AVR platform
These configurations are undergoing the factory test phase and will be soon available
for installation on Ansaldo turbogenerators. The qualification also include EMC and
climatic tests carried out on the operating AVR by a separate certification firm, because
configurations b) and c) involve an hardware upgrade, i.e. a new microprocessor and
new interface have been installed to increase the AVR performance.
All the improvements devised during the NIIPT test phase have been applied to these
new versions. As concerns the power plants in Russian Federation, these new regulators
will be available as an alternative to the already certified regulator, but their installation
will be possible only after they successfully repeat the NIIPT test program.