Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

10th CIGRE Southern Africa

Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

Impact of Extremely Low System Loading on Frequency and Small Signal


Stability during the Pandemic

N. GUMEDE1, T. MODISANE1, B. SIAVHE1, P. DAVEL1, T. MACHABE1, M. NTUSI1


Eskom Holdings SOC1
South Africa

SUMMARY

On 15 March 2020, the president of the Republic of South Africa declared COVID-19
a national disaster. Further, on 23 March 2020, the president announced a raft of
measures to combat the pandemic. These included a 21-day national lockdown
effective from midnight, 26 March 2020 until midnight, 16 April 2020. This lockdown
was later extended by a further 2 weeks. A significant drop in load was visible
following a reduction and in some cases complete shutdown of business activity. In
response to this, a number of coal fired generating units and a nuclear unit were
taken offline and this reduction in the number of online units meant the grid had
significantly lower inertia and was prone to frequency excursions following
disturbances. It was therefore crucial to investigate the impact of the severe
reduction in system loading on power system stability to ensure that the reliability of
the interconnected power system was not compromised. This paper explains
methodologies employed and detailed analysis conducted for frequency and small
signal stability. A dynamic simulation study was conducted to assess the impact on
frequency stability using the detailed Eskom network. The power system
performance during different contingencies (i.e. generator trip, load trip) was
analysed and findings were recorded. A study to ascertain the position and possible
movement of the known system oscillatory modes was conducted using the state of
the art Wide Area Monitoring System. A number of events were analysed and the
data analysis assisted in determining the possible shifting of the known oscillatory
modes. This was required in order to identify and understand any risks that the
Eskom grid may be subjected to under the altered operating conditions.

KEYWORDS

Eskom, Frequency Stability, Small Signal Stability, Power System Stability, Wide
Area Monitoring, Pandemic

GumedeNk@eskom.co.za
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

1 INTRODUCTION
On 15 March 2020, the president of the Republic of South Africa declared COVID-19
as a national disaster. A national lockdown was imposed from the 26th of March.
This was initially supposed to be in effect for two weeks, but this was later extended
by a further 2 week period. A significant drop in load was visible following a
reduction and in some cases complete shutdown of business activity. In response to
this a number of coal fired generating units and a nuclear unit were taken offline and
this reduction in the number of online units meant the grid had significantly lower
inertia and was prone to frequency excursions following disturbances. The reduction
in load also led to de-loading of the online generators, which reduces the system’s
ability to further reduce output in response to load-loss events and led to average
frequencies being above nominal for long periods. To illustrate the extent of the load
reduction, the total residual demand profile vs forecast is plotted over the lockdown
compared to a week before lockdown i.e. from 15 March 2020. The load profile
during lockdown was not only much lower than the “normal” profile for this time of
year, but was also below the traditional low load periods around Christmas and New
Year (see figure 2).

Figure 1 Eskom’s residual hourly demand profile

Figure 2 Eskom’s residual monthly demand profile (2019 vs 2020)

2
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

It was therefore crucial to investigate the impact of the severe reduction in system
loading on power system stability to ensure that the reliability of the interconnected
power system was not compromised. Frequency and small signal stability analysis
was conducted in line with [1] taking into account South African Grid Code
requirements stipulated in [2].

2 STUDY METHODOLOGY
2.1 Frequency stability
The network was configured to align with the scenario of interest and assumptions
described in section 3. The credible worst contingency for load and generator were
identified as indicated in section 3. A dynamic simulation study was conducted to
assess the impact on frequency stability using the detailed Eskom network. During
the study, the relevant contingencies (credible single contingency and credible
multiple contingency) were applied to the network and system performance was
observed over the first 10 seconds. The power system performance was analysed
and findings were recorded.
2.2 Small signal stability
The state of the art Wide Area Monitoring System (WAMS) was used to ascertain the
position and possible movement of the oscillatory modes in the frequency domain.
The WAMS uses Phasor Measurement Units (PMUs) located at strategic locations
on the transmission network, as shown in Figure 3, to monitor the dynamic stability of
the power system at a resolution of 20 ms

Figure 3 PMU locations on the transmission network

3 SIMULATION ASSUMPTIONS
The assumptions applicable to the load trip studies were as following:
 The official System Operator casefile developed and tested by power system
engineers was used in the study
 The generation mix was based on the typical dispatch during light loading

3
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa
 Online generators were not allowed to operate below declared minimum
stable generation level when proving mandatory governing down;
 A realistic number of conventional units had mandatory governing active;
 As the study considered a minimum load condition, corresponding to late
night/early morning loading, PV plants were switched off and only wind plants
were in operation and all had mandatory frequency response;
 System load was 15GW, 20% constant current load and 80% constant
impedance load to cater for industry partial shutdown. Under normal
conditions 100% constant current represents system performance accurately;
 The credible single contingency was represented by trip 2 x 447 MW smelter
loads;
 The credible multiple contingency was not studied;
 Online inverter-based generation capacity approximately 1300 MW, based on
typical wind generation during light loading (experienced at night).

The assumptions applicable to the generator trip studies were as following:


 The official System Operator casefile developed and tested by power system
engineers was used in the study
 The generation mix was based on the typical dispatch during light loading
 First 3 Under frequency load shedding (UFLS) stages were applicable;
 UFLS operating time = 300 ms relay operating (delay) time + 100 ms breaker
operating time;
 Instantaneous Reserve (IR) provision of approximately 250 MW;
 System load was 15GW, 20% constant current load and 80% constant
impedance load to cater for industry partial shutdown. Under normal
conditions 100% constant current represents system performance accurately;
 The credible single contingency was a trip of 930 MW of generation;
 The credible multiple contingency was represented by tripping of Apollo high
voltage direct current (HVDC) lines imports from 1200 MW;
 Online inverter-based generation capacity approximately 1300 MW, based on
typical wind generation during light loading (experienced at night).

There were no simulations conducted for small signal stability analysis therefore
assumptions are not applicable.

4 STUDY RESULTS
4.1 Performance of different types of generators during load trip
Most online generators were expected to be operated close to minimum stable
generation level during low load conditions. Hence, there was a need to analyse the
performance of different machines. When generators are running very close to
minimum stable generation level (minimum generation) they have a limited capacity
available to respond downwards to frequency deviations. The typical governing
response profiles are shown below:

4
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

Figure 4: Non-governing and governing down while close to minimum stable generation level
Figure 4 (left) above shows how a unit with governor functionality switched off
responded to the frequency deviation. At 0 seconds, the load trip contingency
occurred. Initially a reduction of generator output was observed. This was not due to
governing; but it is an inherent characteristic of a synchronous generator behaviour
during this type of disturbance. After 3 seconds, the generator output returns to the
pre-disturbance level. The generator model was configured such that a unit cannot
operate below the minimum stable level. As a result, generators that were close to
minimum stable level before the contingency had a limited capacity to respond.
Figure 4 (right) above shows the typical response when the unit was close to
minimum stable level. The generator output can go below minimum stable level for
few seconds after the disturbance due to the generator natural response to frequency
deviation. However, the output is then maintained at minimum stable capacity as far
as possible after the transient period.

Figure 5 Conventional plant and inverter-based governing down performance


Figure 5 (left) shows the typical response of a generator with enough capacity to
support the grid. Initially the generator output changes due to the natural response to
frequency deviation. A few seconds later, the generator output is reduced further to
provide the required support to the grid. This is due to the generator governing
response. Figure 5 (right) shows the typical response of the wind plant during
frequency disturbance. The wind plants do not exhibit the same natural response as
synchronous generators following the loss of load. They maintain the pre-
disturbance generation output level until the mandatory frequency response trigger
condition is met and then respond accordingly.

5
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa
4.2 Frequency stability during load trip
Firstly, the single credible contingency was applied on a 15 GW load system when
the system frequency was at nominal (50 Hz). Figure 6 shows that when the load trip
was experienced under these conditions the frequency reached 51 Hz after 10
seconds. The maximum frequency recorded was 51.06 Hz and the frequency was
recovered to below 51 Hz within 25 seconds. However, it settled very close to 51 Hz.

Figure 6 System performance when pre disturbance frequency is 50 Hz


In practice the system frequency was not maintained at 50 Hz all the time during hard
lockdown, therefore, it was important to study a load trip from slightly above nominal
frequency. In the second scenario the single credible contingency was applied on a
15 GW load system when system frequency was at 50.15 Hz, which corresponds to
the upper limit of the frequency control dead-band. Figure 7 shows the results, the
frequency reached 51 Hz in 8 seconds and the maximum frequency observed was
51.17 Hz.

Figure 7 System performance when pre disturbance frequency is 50.15 Hz


The primary frequency control action did not recover frequency to below 51 Hz, this
means secondary and tertiary control resources must be dispatched to recover
frequency. The study results show that a large load trip can cause frequency

6
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa
deviation beyond the continuous operation region (49 Hz – 51 Hz). This presents a
potential risk of reaching generator protection trip levels (frequency above 51.5 Hz).
4.3 Frequency stability during generator trip
In order to protect the system from operating in the under frequency region the
System Operator (SO) implements various specialized protection schemes (SPS).
One of these SPSs is the under frequency load shedding scheme (UFLS). The first
three stages of UFLS were relevant on the study and details are as follows:
Table 1 : UFLS stages

Online generators were operated close to minimum stable level most of the time due
to low system load. Therefore they were expected to have enough capacity to
respond to generator trips. However, wind and solar power plants are not contracted
and there is no mandatory requirement to provide grid support during low frequency
disturbances. In the first scenario the single credible contingency was applied on the
network with a loading of 15 GW. The system response is shown in Figure 8. It can
be seen that the Grid Code limit of 49.5 Hz stipulated in [2] was breached. The
frequency response activated two UFLS stages.

Figure 8 15 GW system performance following credible single contingency

In the second scenario the credible multiple contingency was applied to the system
with a loading of 15 GW. The system response is shown Figure 9. It can be seen
that the Grid Code limit of 49.0 Hz stipulated in [2] was breached and three UFLS
stages were activated.

7
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

Figure 9 15 GW system performance following credible multiple contingency


The study results show that a single or multiple credible contingency can cause
frequency deviation beyond Grid Code limits. This presents a risk of activating two
UFLS stages during single contingency and three UFLS stages during multiple
contingencies.
4.4 Small signal stability
A number of small signal events were analysed using the Phasor Measurement Units
(PMU’s) located at the northern and the southern generation power pools on the
transmission grid. The locations of PMU’s are particularly important in observing
dominant oscillatory modes which are prevalent on the network as shown in Figure
10. Historical events that occurred between 2002 and 2011 are shown in Table 2.
Table 2 : Events analysed
Incident Date Power Prevalent Inter Area Modes Network
[MW] [Hz] Loading [MW]
Gen trip event 1 2002/04/20 467 0.3 0.6 20549.90
Load trip event 1 2006/04/20 447 0.3 0.6 23693.50
Gen trip event 2 2006/04/20 467 0.3 0.5 0.7 18211.30
Gen trip event 3 2008/04/20 320 0.3 0.45 15002.90
Gen trip event 4 2008/04/20 287 0.3 0.65 22151.60
Gen trip event 5 2011/08/20 350 0.3 0.6 15355.30

The historical small signal events in Table 2 show the prevalent inter-area oscillatory
modes during low network loading conditions due to generator trip or load loss

8
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa
events. These are the known oscillatory modes of predominantly 0.3 Hz and 0.6 Hz
in nature.

5 SYSTEM PERFORMANCE DURING HARD LOCKDOWN


The frequency stability simulations were conducted to understand how the system
will perform following credible single contingency and credible multiple unit trip as
defined by the Grid Code. The largest generator trip experienced during extreme
light loading was the loss of 726 MW, which occurred when system loading was just
below 20 GW and system frequency was within the dead-band (49.85 Hz – 50.15 Hz)
as shown in Figure 10 below.

Figure 10 System performance during the loss of 726 MW generation


As seen in Figure 10, the system was able to deal with a generator trip from 726 MW
without violating Grid Code frequency limits. There were no reportable load trips or
generator trips greater than 930 MW.

The few events analysed for small signal stability during the low demand period are
shown in Table 3. These events showed no evidence of the known oscillatory modes
having changed in any significant way. A statistical view of the modes of oscillation
in events shows that all the observed modes had significant positive damping with
the minimum damping ratio above 4% for the 0.3 Hz mode. This is an acceptable
limit because the current alarm limit for the 0.2 - 0.8 Hz mode is 2% damping ratio.

Table 3 : SAPP tie-line histogram statistics

9
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

Figure 11 System frequency oscillatory modes


The first event occurred on the 29th March 2020 at 09:06:22. The histogram plot in
Figure 11 shows the most prevalent modes of oscillation during the event. Figure 11
shows the 0.3 Hz mode being the dominant oscillation mode on the system. There is
a small component of the 0.55 Hz mode occurring as well as more localized modes
above 0.8 Hz.

6 CONCLUSIONS
The system loading influences power system frequency response. The load model
assumption was based on the fact that industries in the country were on partial
lockdown. The study results show that a large load trip can cause frequency
deviation beyond the continuous operation region (49 Hz – 51 Hz). This presents a
potential risk of reaching generator protection trip levels (frequency above 51.5 Hz).
The study results also showed that a single or multiple credible generation
contingency can cause frequency deviation beyond Grid Code limits. This presents a
risk of activating two UFLS stages during single contingency and three UFLS stages
during multiple contingency events. The system measurements show that the largest
generator trip experienced during extreme light loading was the loss of 726 MW,
which is less than the credible worst case single contingency. The power system
was able to deal with the loss of the 726 MW without violating Grid Code frequency
limits. There were no reportable load trips or generator trips greater than 930 MW. A
data analysis exercise was conducted through the use of WAMS (Wide Area
Monitoring System) to evaluate the possible shifting of the known oscillatory modes.
The analysis revealed that there was no evidence of significant changes in the
existing modes (0.3 and 0.6 Hz) of oscillation due to the low power demand. This
was evident from the comparison of the WAMS data during the lockdown period and
the historical data of the oscillatory modes. Furthermore, the observed modes of
oscillation were well damped in accordance to the historical data had no real impact
on the stability of the power system.

10
10th CIGRE Southern Africa
Regional Conference
2nd – 4th November 2021
Johannesburg, South Africa

BIBLIOGRAPHY
[1] P. Kundur, Power System Stability and Control, (McGraw-Hill: New York, 1994).
[2] South African Grid Code version 10, 2019
[3] B. Pal and B. Chaudhuri, Robust Control in Power Systems (Springer Inc. New
York, 2005).
[4] Komkrit Prasertwong, Mithulananthan Nadarajah, Devbratta Thakur,
“Understanding Low Frequency Oscillations in Power Systems”. International
Journal of Electrical Engineering Education · July 2010.

11

You might also like